refactor(aliases): migrate to pattern-based model alias resolution

Move from hardcoded exact aliases to pattern-based canonicalization:

- Populate PATTERN_ALIAS_RULES with regex patterns for:
  - Claude thinking variants (claude-opus-4-6-thinking → claude-opus-4-6)
  - Gemini tier suffixes (gemini-3.1-pro-{high,low} → gemini-3.1-pro)
- Add stripProviderPrefixForAliasLookup() for provider-prefixed models
  (anthropic/claude-sonnet-4-6 → claude-sonnet-4-6 for capability lookup)
- Preserve requestedModelID (with prefix) for API transport
- Reduce EXACT_ALIAS_RULES to exceptional cases only
  (gemini-3-pro-{high,low} → gemini-3-pro-preview)
- Comprehensive test coverage for patterns, prefix stripping, negatives

Addresses Discussion #2835 (pattern matching architecture)
Related to PR #2834 (alias guardrails)

41 targeted tests pass, 4467 full suite tests pass, tsc clean.

.github/workflows/publish-platform.yml

@@ -59,20 +59,39 @@ jobs:
       - name: Check if already published
         id: check
         run: |
-          PKG_NAME="oh-my-opencode-${{ matrix.platform }}"
           VERSION="${{ inputs.version }}"
-          STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/${PKG_NAME}/${VERSION}")
-          # Convert platform name for output (replace - with _)
           PLATFORM_KEY="${{ matrix.platform }}"
           PLATFORM_KEY="${PLATFORM_KEY//-/_}"
-          if [ "$STATUS" = "200" ]; then
+          
+          # Check oh-my-opencode
+          OC_STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/oh-my-opencode-${{ matrix.platform }}/${VERSION}")
+          # Check oh-my-openagent
+          OA_STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/oh-my-openagent-${{ matrix.platform }}/${VERSION}")
+          
+          echo "oh-my-opencode-${{ matrix.platform }}@${VERSION}: ${OC_STATUS}"
+          echo "oh-my-openagent-${{ matrix.platform }}@${VERSION}: ${OA_STATUS}"
+          
+          if [ "$OC_STATUS" = "200" ]; then
+            echo "skip_opencode=true" >> $GITHUB_OUTPUT
+            echo "✓ oh-my-opencode-${{ matrix.platform }}@${VERSION} already published"
+          else
+            echo "skip_opencode=false" >> $GITHUB_OUTPUT
+            echo "→ oh-my-opencode-${{ matrix.platform }}@${VERSION} needs publishing"
+          fi
+          
+          if [ "$OA_STATUS" = "200" ]; then
+            echo "skip_openagent=true" >> $GITHUB_OUTPUT
+            echo "✓ oh-my-openagent-${{ matrix.platform }}@${VERSION} already published"
+          else
+            echo "skip_openagent=false" >> $GITHUB_OUTPUT
+            echo "→ oh-my-openagent-${{ matrix.platform }}@${VERSION} needs publishing"
+          fi
+          
+          # Skip build only if BOTH are already published
+          if [ "$OC_STATUS" = "200" ] && [ "$OA_STATUS" = "200" ]; then
             echo "skip=true" >> $GITHUB_OUTPUT
-            echo "skip_${PLATFORM_KEY}=true" >> $GITHUB_OUTPUT
-            echo "✓ ${PKG_NAME}@${VERSION} already published"
           else
             echo "skip=false" >> $GITHUB_OUTPUT
-            echo "skip_${PLATFORM_KEY}=false" >> $GITHUB_OUTPUT
-            echo "→ ${PKG_NAME}@${VERSION} needs publishing"
           fi
 
       - name: Update version in package.json
@@ -82,6 +101,11 @@ jobs:
           cd packages/${{ matrix.platform }}
           jq --arg v "$VERSION" '.version = $v' package.json > tmp.json && mv tmp.json package.json
 
+      - name: Set root package version
+        if: steps.check.outputs.skip != 'true'
+        run: |
+          jq --arg v "${{ inputs.version }}" '.version = $v' package.json > tmp.json && mv tmp.json package.json
+
       - name: Pre-download baseline compile target
         if: steps.check.outputs.skip != 'true' && endsWith(matrix.platform, '-baseline')
         shell: bash
@@ -192,11 +216,6 @@ jobs:
           retention-days: 1
           if-no-files-found: error
 
-  # =============================================================================
-  # Job 2: Publish all platforms (oh-my-opencode + oh-my-openagent)
-  # - Runs on ubuntu-latest for ALL platforms (just downloading artifacts)
-  # - Uses NODE_AUTH_TOKEN for auth + OIDC for provenance attestation
-  # =============================================================================
   publish:
     needs: build
     if: always() && !cancelled()
@@ -207,37 +226,38 @@ jobs:
       matrix:
         platform: [darwin-arm64, darwin-x64, darwin-x64-baseline, linux-x64, linux-x64-baseline, linux-arm64, linux-x64-musl, linux-x64-musl-baseline, linux-arm64-musl, windows-x64, windows-x64-baseline]
     steps:
-      - name: Check if oh-my-opencode already published
+      - name: Check if already published
         id: check
         run: |
-          PKG_NAME="oh-my-opencode-${{ matrix.platform }}"
           VERSION="${{ inputs.version }}"
-          STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/${PKG_NAME}/${VERSION}")
+          
-          if [ "$STATUS" = "200" ]; then
+          OC_STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/oh-my-opencode-${{ matrix.platform }}/${VERSION}")
-            echo "skip=true" >> $GITHUB_OUTPUT
+          OA_STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/oh-my-openagent-${{ matrix.platform }}/${VERSION}")
-            echo "✓ ${PKG_NAME}@${VERSION} already published, skipping"
+          
+          if [ "$OC_STATUS" = "200" ]; then
+            echo "skip_opencode=true" >> $GITHUB_OUTPUT
+            echo "✓ oh-my-opencode-${{ matrix.platform }}@${VERSION} already published"
           else
-            echo "skip=false" >> $GITHUB_OUTPUT
+            echo "skip_opencode=false" >> $GITHUB_OUTPUT
-            echo "→ ${PKG_NAME}@${VERSION} will be published"
           fi
           
-      - name: Check if oh-my-openagent already published
+          if [ "$OA_STATUS" = "200" ]; then
-        id: check-openagent
+            echo "skip_openagent=true" >> $GITHUB_OUTPUT
-        run: |
+            echo "✓ oh-my-openagent-${{ matrix.platform }}@${VERSION} already published"
-          PKG_NAME="oh-my-openagent-${{ matrix.platform }}"
-          VERSION="${{ inputs.version }}"
-          STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/${PKG_NAME}/${VERSION}")
-          if [ "$STATUS" = "200" ]; then
-            echo "skip=true" >> $GITHUB_OUTPUT
-            echo "✓ ${PKG_NAME}@${VERSION} already published, skipping"
           else
-            echo "skip=false" >> $GITHUB_OUTPUT
+            echo "skip_openagent=false" >> $GITHUB_OUTPUT
-            echo "→ ${PKG_NAME}@${VERSION} will be published"
+          fi
+          
+          # Need artifact if either package needs publishing
+          if [ "$OC_STATUS" = "200" ] && [ "$OA_STATUS" = "200" ]; then
+            echo "skip_all=true" >> $GITHUB_OUTPUT
+          else
+            echo "skip_all=false" >> $GITHUB_OUTPUT
           fi
 
       - name: Download artifact
         id: download
-        if: steps.check.outputs.skip != 'true' || steps.check-openagent.outputs.skip != 'true'
+        if: steps.check.outputs.skip_all != 'true'
         continue-on-error: true
         uses: actions/download-artifact@v4
         with:
@@ -245,7 +265,7 @@ jobs:
           path: .
 
       - name: Extract artifact
-        if: (steps.check.outputs.skip != 'true' || steps.check-openagent.outputs.skip != 'true') && steps.download.outcome == 'success'
+        if: steps.check.outputs.skip_all != 'true' && steps.download.outcome == 'success'
         run: |
           PLATFORM="${{ matrix.platform }}"
           mkdir -p packages/${PLATFORM}
@@ -261,13 +281,13 @@ jobs:
           ls -la packages/${PLATFORM}/bin/
 
       - uses: actions/setup-node@v4
-        if: (steps.check.outputs.skip != 'true' || steps.check-openagent.outputs.skip != 'true') && steps.download.outcome == 'success'
+        if: steps.check.outputs.skip_all != 'true' && steps.download.outcome == 'success'
         with:
           node-version: "24"
           registry-url: "https://registry.npmjs.org"
 
-      - name: Publish ${{ matrix.platform }}
+      - name: Publish oh-my-opencode-${{ matrix.platform }}
-        if: steps.check.outputs.skip != 'true' && steps.download.outcome == 'success'
+        if: steps.check.outputs.skip_opencode != 'true' && steps.download.outcome == 'success'
         run: |
           cd packages/${{ matrix.platform }}
           
@@ -283,7 +303,7 @@ jobs:
         timeout-minutes: 15
 
       - name: Publish oh-my-openagent-${{ matrix.platform }}
-        if: steps.check-openagent.outputs.skip != 'true' && steps.download.outcome == 'success'
+        if: steps.check.outputs.skip_openagent != 'true' && steps.download.outcome == 'success'
         run: |
           cd packages/${{ matrix.platform }}
           

.github/workflows/publish.yml

@@ -57,6 +57,7 @@ jobs:
           bun test src/cli/doctor/format-default.test.ts
           bun test src/tools/call-omo-agent/sync-executor.test.ts
           bun test src/tools/call-omo-agent/session-creator.test.ts
+          bun test src/tools/session-manager
           bun test src/features/opencode-skill-loader/loader.test.ts
           bun test src/hooks/anthropic-context-window-limit-recovery/recovery-hook.test.ts
           bun test src/hooks/anthropic-context-window-limit-recovery/executor.test.ts
@@ -66,9 +67,8 @@ jobs:
           # Enumerate subdirectories/files explicitly to EXCLUDE mock-heavy files
           # that were already run in isolation above.
           # Excluded from src/cli: doctor/formatter.test.ts, doctor/format-default.test.ts
-          # Excluded from src/tools: call-omo-agent/sync-executor.test.ts, call-omo-agent/session-creator.test.ts
+          # Excluded from src/tools: call-omo-agent/sync-executor.test.ts, call-omo-agent/session-creator.test.ts, session-manager (all)
           # Excluded from src/hooks/anthropic-context-window-limit-recovery: recovery-hook.test.ts, executor.test.ts
-          # Excluded from src/tools: call-omo-agent/sync-executor.test.ts, call-omo-agent/session-creator.test.ts
           bun test bin script src/config src/mcp src/index.test.ts \
             src/agents src/shared \
             src/cli/run src/cli/config-manager src/cli/mcp-oauth \
@@ -77,7 +77,7 @@ jobs:
             src/cli/doctor/runner.test.ts src/cli/doctor/checks \
             src/tools/ast-grep src/tools/background-task src/tools/delegate-task \
             src/tools/glob src/tools/grep src/tools/interactive-bash \
-            src/tools/look-at src/tools/lsp src/tools/session-manager \
+            src/tools/look-at src/tools/lsp \
             src/tools/skill src/tools/skill-mcp src/tools/slashcommand src/tools/task \
             src/tools/call-omo-agent/background-agent-executor.test.ts \
             src/tools/call-omo-agent/background-executor.test.ts \
@@ -215,40 +215,48 @@ jobs:
         env:
           NODE_AUTH_TOKEN: ${{ secrets.NODE_AUTH_TOKEN }}
           NPM_CONFIG_PROVENANCE: true
-      - name: Publish oh-my-openagent
-        if: steps.check.outputs.skip != 'true'
-        run: |
-          # Update package name to oh-my-openagent
-          jq '.name = "oh-my-openagent"' package.json > tmp.json && mv tmp.json package.json
 
-          # Update optionalDependencies to use oh-my-openagent naming
+      - name: Check if oh-my-openagent already published
-          jq '.optionalDependencies = {
+        id: check-openagent
-            "oh-my-openagent-darwin-arm64": "${{ steps.version.outputs.version }}",
+        run: |
-            "oh-my-openagent-darwin-x64": "${{ steps.version.outputs.version }}",
+          VERSION="${{ steps.version.outputs.version }}"
-            "oh-my-openagent-darwin-x64-baseline": "${{ steps.version.outputs.version }}",
+          STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/oh-my-openagent/${VERSION}")
-            "oh-my-openagent-linux-arm64": "${{ steps.version.outputs.version }}",
+          if [ "$STATUS" = "200" ]; then
-            "oh-my-openagent-linux-arm64-musl": "${{ steps.version.outputs.version }}",
+            echo "skip=true" >> $GITHUB_OUTPUT
-            "oh-my-openagent-linux-x64": "${{ steps.version.outputs.version }}",
+            echo "✓ oh-my-openagent@${VERSION} already published"
-            "oh-my-openagent-linux-x64-baseline": "${{ steps.version.outputs.version }}",
+          else
-            "oh-my-openagent-linux-x64-musl": "${{ steps.version.outputs.version }}",
+            echo "skip=false" >> $GITHUB_OUTPUT
-            "oh-my-openagent-linux-x64-musl-baseline": "${{ steps.version.outputs.version }}",
+          fi
-            "oh-my-openagent-windows-x64": "${{ steps.version.outputs.version }}",
+
-            "oh-my-openagent-windows-x64-baseline": "${{ steps.version.outputs.version }}"
+      - name: Publish oh-my-openagent
-          }' package.json > tmp.json && mv tmp.json package.json
+        if: steps.check-openagent.outputs.skip != 'true'
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          
+          # Update package name, version, and optionalDependencies for oh-my-openagent
+          jq --arg v "$VERSION" '
+            .name = "oh-my-openagent" |
+            .version = $v |
+            .optionalDependencies = (
+              .optionalDependencies | to_entries |
+              map(.key = (.key | sub("^oh-my-opencode-"; "oh-my-openagent-")) | .value = $v) |
+              from_entries
+            )
+          ' package.json > tmp.json && mv tmp.json package.json
           
           TAG_ARG=""
           if [ -n "${{ steps.version.outputs.dist_tag }}" ]; then
             TAG_ARG="--tag ${{ steps.version.outputs.dist_tag }}"
           fi
-          npm publish --access public --provenance $TAG_ARG || echo "oh-my-openagent publish may have failed (package may already exist)"
+          npm publish --access public --provenance $TAG_ARG || echo "::warning::oh-my-openagent publish failed"
         env:
           NODE_AUTH_TOKEN: ${{ secrets.NODE_AUTH_TOKEN }}
           NPM_CONFIG_PROVENANCE: true
+
       - name: Restore package.json
-        if: steps.check.outputs.skip != 'true'
+        if: steps.check-openagent.outputs.skip != 'true'
         run: |
-          # Restore original package name
+          git checkout -- package.json
-          jq '.name = "oh-my-opencode"' package.json > tmp.json && mv tmp.json package.json
 
   trigger-platform:
     runs-on: ubuntu-latest

.github/workflows/refresh-model-capabilities.yml

@@ -0,0 +1,46 @@
+name: Refresh Model Capabilities
+
+on:
+  schedule:
+    - cron: "17 4 * * 1"
+  workflow_dispatch:
+
+permissions:
+  contents: write
+  pull-requests: write
+
+jobs:
+  refresh:
+    runs-on: ubuntu-latest
+    if: github.repository == 'code-yeongyu/oh-my-openagent'
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: oven-sh/setup-bun@v2
+        with:
+          bun-version: latest
+
+      - name: Install dependencies
+        run: bun install
+        env:
+          BUN_INSTALL_ALLOW_SCRIPTS: "@ast-grep/napi"
+
+      - name: Refresh bundled model capabilities snapshot
+        run: bun run build:model-capabilities
+
+      - name: Validate capability guardrails
+        run: bun run test:model-capabilities
+
+      - name: Create refresh pull request
+        uses: peter-evans/create-pull-request@v7
+        with:
+          commit-message: "chore: refresh model capabilities snapshot"
+          title: "chore: refresh model capabilities snapshot"
+          body: |
+            Automated refresh of `src/generated/model-capabilities.generated.json` from `https://models.dev/api.json`.
+
+            This keeps the bundled capability snapshot aligned with upstream model metadata without relying on manual refreshes.
+          branch: automation/refresh-model-capabilities
+          delete-branch: true
+          labels: |
+            maintenance

.gitignore

@@ -9,6 +9,7 @@ dist/
 # Platform binaries (built, not committed)
 packages/*/bin/oh-my-opencode
 packages/*/bin/oh-my-opencode.exe
+packages/*/bin/*.map
 
 # IDE
 .idea/
@@ -35,3 +36,4 @@ test-injection/
 notepad.md
 oauth-success.html
 *.bun-build
+.omx/

.opencode/skills/github-triage/SKILL.md

@@ -1,105 +1,229 @@
 ---
 name: github-triage
-description: "Unified GitHub triage for issues AND PRs. 1 item = 1 background task (category: free). Issues: answer questions from codebase, analyze bugs. PRs: review bugfixes, merge safe ones. All parallel, all background. Triggers: 'triage', 'triage issues', 'triage PRs', 'github triage'."
+description: "Read-only GitHub triage for issues AND PRs. 1 item = 1 background task (category: quick). Analyzes all open items and writes evidence-backed reports to /tmp/{datetime}/. Every claim requires a GitHub permalink as proof. NEVER takes any action on GitHub - no comments, no merges, no closes, no labels. Reports only. Triggers: 'triage', 'triage issues', 'triage PRs', 'github triage'."
 ---
 
-# GitHub Triage — Unified Issue & PR Processor
+# GitHub Triage - Read-Only Analyzer
 
 <role>
-You are a GitHub triage orchestrator. You fetch all open issues and PRs, classify each one, then spawn exactly 1 background subagent per item using `category="free"`. Each subagent analyzes its item, takes action (comment/close/merge/report), and records results via TaskCreate.
+Read-only GitHub triage orchestrator. Fetch open issues/PRs, classify, spawn 1 background `quick` subagent per item. Each subagent analyzes and writes a report file. ZERO GitHub mutations.
 </role>
 
----
+## Architecture
 
-## ARCHITECTURE
+**1 ISSUE/PR = 1 `task_create` = 1 `quick` SUBAGENT (background). NO EXCEPTIONS.**
-
-```
-1 issue or PR = 1 TaskCreate = 1 task(category="free", run_in_background=true)
-```
 
 | Rule | Value |
 |------|-------|
-| Category for ALL subagents | `free` |
+| Category | `quick` |
-| Execution mode | `run_in_background=true` |
+| Execution | `run_in_background=true` |
-| Parallelism | ALL items launched simultaneously |
+| Parallelism | ALL items simultaneously |
-| Result tracking | Each subagent calls `TaskCreate` with its findings |
+| Tracking | `task_create` per item |
-| Result collection | `background_output()` polling loop |
+| Output | `/tmp/{YYYYMMDD-HHmmss}/issue-{N}.md` or `pr-{N}.md` |
 
 ---
 
-## PHASE 1: FETCH ALL OPEN ITEMS
+## Zero-Action Policy (ABSOLUTE)
 
-<fetch>
+<zero_action>
-Run these commands to collect data. Use the bundled script if available, otherwise fall back to gh CLI.
+Subagents MUST NEVER run ANY command that writes or mutates GitHub state.
+
+**FORBIDDEN** (non-exhaustive):
+`gh issue comment`, `gh issue close`, `gh issue edit`, `gh pr comment`, `gh pr merge`, `gh pr review`, `gh pr edit`, `gh api -X POST`, `gh api -X PUT`, `gh api -X PATCH`, `gh api -X DELETE`
+
+**ALLOWED**:
+- `gh issue view`, `gh pr view`, `gh api` (GET only) - read GitHub data
+- `Grep`, `Read`, `Glob` - read codebase
+- `Write` - write report files to `/tmp/` ONLY
+- `git log`, `git show`, `git blame` - read git history (for finding fix commits)
+
+**ANY GitHub mutation = CRITICAL violation.**
+</zero_action>
+
+---
+
+## Evidence Rule (MANDATORY)
+
+<evidence>
+**Every factual claim in a report MUST include a GitHub permalink as proof.**
+
+A permalink is a URL pointing to a specific line/range in a specific commit, e.g.:
+`https://github.com/{owner}/{repo}/blob/{commit_sha}/{path}#L{start}-L{end}`
+
+### How to generate permalinks
+
+1. Find the relevant file and line(s) via Grep/Read.
+2. Get the current commit SHA: `git rev-parse HEAD`
+3. Construct: `https://github.com/{REPO}/blob/{SHA}/{filepath}#L{line}` (or `#L{start}-L{end}` for ranges)
+
+### Rules
+
+- **No permalink = no claim.** If you cannot back a statement with a permalink, state "No evidence found" instead.
+- Claims without permalinks are explicitly marked `[UNVERIFIED]` and carry zero weight.
+- Permalinks to `main`/`master`/`dev` branches are NOT acceptable - use commit SHAs only.
+- For bug analysis: permalink to the problematic code. For fix verification: permalink to the fixing commit diff.
+</evidence>
+
+---
+
+## Phase 0: Setup
 
 ```bash
 REPO=$(gh repo view --json nameWithOwner -q .nameWithOwner)
-
+REPORT_DIR="/tmp/$(date +%Y%m%d-%H%M%S)"
-# Issues: all open
+mkdir -p "$REPORT_DIR"
-gh issue list --repo $REPO --state open --limit 500 \
+COMMIT_SHA=$(git rev-parse HEAD)
-  --json number,title,state,createdAt,updatedAt,labels,author,body,comments
-
-# PRs: all open
-gh pr list --repo $REPO --state open --limit 500 \
-  --json number,title,state,createdAt,updatedAt,labels,author,body,headRefName,baseRefName,isDraft,mergeable,reviewDecision,statusCheckRollup
 ```
 
-If either returns exactly 500 results, paginate using `--search "created:<LAST_CREATED_AT"` until exhausted.
+Pass `REPO`, `REPORT_DIR`, and `COMMIT_SHA` to every subagent.
-</fetch>
 
 ---
 
-## PHASE 2: CLASSIFY EACH ITEM
+---
 
-For each item, determine its type based on title, labels, and body content:
+## Phase 1: Fetch All Open Items (CORRECTED)
 
-<classification>
+**IMPORTANT:** `body` and `comments` fields may contain control characters that break jq parsing. Fetch basic metadata first, then fetch full details per-item in subagents.
 
-### Issues
+```bash
+# Step 1: Fetch basic metadata (without body/comments to avoid JSON parsing issues)
+ISSUES_LIST=$(gh issue list --repo $REPO --state open --limit 500 \
+  --json number,title,labels,author,createdAt)
+ISSUE_COUNT=$(echo "$ISSUES_LIST" | jq length)
 
-| Type | Detection | Action Path |
+# Paginate if needed
-|------|-----------|-------------|
+if [ "$ISSUE_COUNT" -eq 500 ]; then
-| `ISSUE_QUESTION` | Title contains `[Question]`, `[Discussion]`, `?`, or body is asking "how to" / "why does" / "is it possible" | SUBAGENT_ISSUE_QUESTION |
+  LAST_DATE=$(echo "$ISSUES_LIST" | jq -r '.[-1].createdAt')
-| `ISSUE_BUG` | Title contains `[Bug]`, `Bug:`, body describes unexpected behavior, error messages, stack traces | SUBAGENT_ISSUE_BUG |
+  while true; do
-| `ISSUE_FEATURE` | Title contains `[Feature]`, `[RFE]`, `[Enhancement]`, `Feature Request`, `Proposal` | SUBAGENT_ISSUE_FEATURE |
+    PAGE=$(gh issue list --repo $REPO --state open --limit 500 \
-| `ISSUE_OTHER` | Anything else | SUBAGENT_ISSUE_OTHER |
+      --search "created:<$LAST_DATE" \
+      --json number,title,labels,author,createdAt)
+    PAGE_COUNT=$(echo "$PAGE" | jq length)
+    [ "$PAGE_COUNT" -eq 0 ] && break
+    ISSUES_LIST=$(echo "$ISSUES_LIST" "$PAGE" | jq -s '.[0] + .[1] | unique_by(.number)')
+    ISSUE_COUNT=$(echo "$ISSUES_LIST" | jq length)
+    [ "$PAGE_COUNT" -lt 500 ] && break
+    LAST_DATE=$(echo "$PAGE" | jq -r '.[-1].createdAt')
+  done
+fi
 
-### PRs
+# Same for PRs
+PRS_LIST=$(gh pr list --repo $REPO --state open --limit 500 \
+  --json number,title,labels,author,headRefName,baseRefName,isDraft,createdAt)
+PR_COUNT=$(echo "$PRS_LIST" | jq length)
 
-| Type | Detection | Action Path |
+if [ "$PR_COUNT" -eq 500 ]; then
-|------|-----------|-------------|
+  LAST_DATE=$(echo "$PRS_LIST" | jq -r '.[-1].createdAt')
-| `PR_BUGFIX` | Title starts with `fix`, `fix:`, `fix(`, branch contains `fix/`, `bugfix/`, or labels include `bug` | SUBAGENT_PR_BUGFIX |
+  while true; do
-| `PR_OTHER` | Everything else (feat, refactor, docs, chore, etc.) | SUBAGENT_PR_OTHER |
+    PAGE=$(gh pr list --repo $REPO --state open --limit 500 \
+      --search "created:<$LAST_DATE" \
+      --json number,title,labels,author,headRefName,baseRefName,isDraft,createdAt)
+    PAGE_COUNT=$(echo "$PAGE" | jq length)
+    [ "$PAGE_COUNT" -eq 0 ] && break
+    PRS_LIST=$(echo "$PRS_LIST" "$PAGE" | jq -s '.[0] + .[1] | unique_by(.number)')
+    PR_COUNT=$(echo "$PRS_LIST" | jq length)
+    [ "$PAGE_COUNT" -lt 500 ] && break
+    LAST_DATE=$(echo "$PAGE" | jq -r '.[-1].createdAt')
+  done
+fi
+
+echo "Total issues: $ISSUE_COUNT, Total PRs: $PR_COUNT"
+```
+
+**LARGE REPOSITORY HANDLING:**
+If total items exceeds 50, you MUST process ALL items. Use the pagination code above to fetch every single open issue and PR.
+**DO NOT** sample or limit to 50 items - process the entire backlog.
+
+Example: If there are 500 open issues, spawn 500 subagents. If there are 1000 open PRs, spawn 1000 subagents.
+
+**Note:** Background task system will queue excess tasks automatically.
 
-</classification>
 
 ---
 
-## PHASE 3: SPAWN 1 BACKGROUND TASK PER ITEM
+## Phase 2: Classify
 
-For EVERY item, create a TaskCreate entry first, then spawn a background task.
+| Type | Detection |
+|------|-----------|
+| `ISSUE_QUESTION` | `[Question]`, `[Discussion]`, `?`, "how to" / "why does" / "is it possible" |
+| `ISSUE_BUG` | `[Bug]`, `Bug:`, error messages, stack traces, unexpected behavior |
+| `ISSUE_FEATURE` | `[Feature]`, `[RFE]`, `[Enhancement]`, `Feature Request`, `Proposal` |
+| `ISSUE_OTHER` | Anything else |
+| `PR_BUGFIX` | Title starts with `fix`, branch contains `fix/`/`bugfix/`, label `bug` |
+| `PR_OTHER` | Everything else |
+
+---
+
+## Phase 3: Spawn Subagents (Individual Tool Calls)
+
+**CRITICAL: Create tasks ONE BY ONE using individual `task_create` tool calls. NEVER batch or script.**
+
+For each item, execute these steps sequentially:
+
+### Step 3.1: Create Task Record
+```typescript
+task_create(
+  subject="Triage: #{number} {title}",
+  description="GitHub {issue|PR} triage analysis - {type}",
+  metadata={"type": "{ISSUE_QUESTION|ISSUE_BUG|ISSUE_FEATURE|ISSUE_OTHER|PR_BUGFIX|PR_OTHER}", "number": {number}}
+)
+```
+
+### Step 3.2: Spawn Analysis Subagent (Background)
+```typescript
+task(
+  category="quick",
+  run_in_background=true,
+  load_skills=[],
+  prompt=SUBAGENT_PROMPT
+)
+```
+
+**ABSOLUTE RULES for Subagents:**
+- **ONLY ANALYZE** - Never take action on GitHub (no comments, merges, closes)
+- **READ-ONLY** - Use tools only for reading code/GitHub data
+- **WRITE REPORT ONLY** - Output goes to `{REPORT_DIR}/{issue|pr}-{number}.md` via Write tool
+- **EVIDENCE REQUIRED** - Every claim must have GitHub permalink as proof
 
 ```
 For each item:
-  1. TaskCreate(subject="Triage: #{number} {title}")
+  1. task_create(subject="Triage: #{number} {title}")
-  2. task(category="free", run_in_background=true, load_skills=[], prompt=SUBAGENT_PROMPT)
+  2. task(category="quick", run_in_background=true, load_skills=[], prompt=SUBAGENT_PROMPT)
   3. Store mapping: item_number -> { task_id, background_task_id }
 ```
 
 ---
 
-## SUBAGENT PROMPT TEMPLATES
+## Subagent Prompts
+
+### Common Preamble (include in ALL subagent prompts)
+
+```
+CONTEXT:
+- Repository: {REPO}
+- Report directory: {REPORT_DIR}
+- Current commit SHA: {COMMIT_SHA}
+
+PERMALINK FORMAT:
+Every factual claim MUST include a permalink: https://github.com/{REPO}/blob/{COMMIT_SHA}/{filepath}#L{start}-L{end}
+No permalink = no claim. Mark unverifiable claims as [UNVERIFIED].
+To get current SHA if needed: git rev-parse HEAD
+
+ABSOLUTE RULES (violating ANY = critical failure):
+- NEVER run gh issue comment, gh issue close, gh issue edit
+- NEVER run gh pr comment, gh pr merge, gh pr review, gh pr edit
+- NEVER run any gh command with -X POST, -X PUT, -X PATCH, -X DELETE
+- NEVER run git checkout, git fetch, git pull, git switch, git worktree
+- Your ONLY writable output: {REPORT_DIR}/{issue|pr}-{number}.md via the Write tool
+```
 
-Each subagent gets an explicit, step-by-step prompt. Free models are limited — leave NOTHING implicit.
 
 ---
 
-### SUBAGENT_ISSUE_QUESTION
+### ISSUE_QUESTION
-
-<issue_question_prompt>
 
 ```
-You are a GitHub issue responder for the repository {REPO}.
+You are analyzing issue #{number} for {REPO}.
 
 ITEM:
 - Issue #{number}: {title}
@@ -107,52 +231,43 @@ ITEM:
 - Body: {body}
 - Comments: {comments_summary}
 
-YOUR JOB:
+TASK:
-1. Read the issue carefully. Understand what the user is asking.
+1. Understand the question.
-2. Search the codebase to find the answer. Use Grep and Read tools.
+2. Search the codebase (Grep, Read) for the answer.
-   - Search for relevant file names, function names, config keys mentioned in the issue.
+3. For every finding, construct a permalink: https://github.com/{REPO}/blob/{COMMIT_SHA}/{path}#L{N}
-   - Read the files you find to understand how the feature works.
+4. Write report to {REPORT_DIR}/issue-{number}.md
-3. Decide: Can you answer this clearly and accurately from the codebase?
 
-IF YES (you found a clear, accurate answer):
+REPORT FORMAT (write this as the file content):
-  Step A: Write a helpful comment. The comment MUST:
-    - Start with exactly: [sisyphus-bot]
-    - Be warm, friendly, and thorough
-    - Include specific file paths and code references
-    - Include code snippets or config examples if helpful
-    - End with "Feel free to reopen if this doesn't resolve your question!"
-  Step B: Post the comment:
-    gh issue comment {number} --repo {REPO} --body "YOUR_COMMENT"
-  Step C: Close the issue:
-    gh issue close {number} --repo {REPO}
-  Step D: Report back with this EXACT format:
-    ACTION: ANSWERED_AND_CLOSED
-    COMMENT_POSTED: yes
-    SUMMARY: [1-2 sentence summary of your answer]
 
-IF NO (not enough info in codebase, or answer is uncertain):
+# Issue #{number}: {title}
-  Report back with:
+**Type:** Question | **Author:** {author} | **Created:** {createdAt}
-    ACTION: NEEDS_MANUAL_ATTENTION
-    REASON: [why you couldn't answer — be specific]
-    PARTIAL_FINDINGS: [what you DID find, if anything]
 
-RULES:
+## Question
-- NEVER guess. Only answer if the codebase clearly supports your answer.
+[1-2 sentence summary]
-- NEVER make up file paths or function names.
+
-- The [sisyphus-bot] prefix is MANDATORY on every comment you post.
+## Findings
-- Be genuinely helpful — imagine you're a senior maintainer who cares about the community.
+[Each finding with permalink proof. Example:]
+- The config is parsed in [`src/config/loader.ts#L42-L58`](https://github.com/{REPO}/blob/{SHA}/src/config/loader.ts#L42-L58)
+
+## Suggested Answer
+[Draft answer with code references and permalinks]
+
+## Confidence: [HIGH | MEDIUM | LOW]
+[Reason. If LOW: what's missing]
+
+## Recommended Action
+[What maintainer should do]
+
+---
+REMEMBER: No permalink = no claim. Every code reference needs a permalink.
 ```
 
-</issue_question_prompt>
-
 ---
 
-### SUBAGENT_ISSUE_BUG
+### ISSUE_BUG
-
-<issue_bug_prompt>
 
 ```
-You are a GitHub bug analyzer for the repository {REPO}.
+You are analyzing bug report #{number} for {REPO}.
 
 ITEM:
 - Issue #{number}: {title}
@@ -160,74 +275,75 @@ ITEM:
 - Body: {body}
 - Comments: {comments_summary}
 
-YOUR JOB:
+TASK:
-1. Read the issue carefully. Understand the reported bug:
+1. Understand: expected behavior, actual behavior, reproduction steps.
-   - What behavior does the user expect?
+2. Search the codebase for relevant code. Trace the logic.
-   - What behavior do they actually see?
+3. Determine verdict: CONFIRMED_BUG, NOT_A_BUG, ALREADY_FIXED, or UNCLEAR.
-   - What steps reproduce it?
+4. For ALREADY_FIXED: find the fixing commit using git log/git blame. Include the commit SHA and what changed.
-2. Search the codebase for the relevant code. Use Grep and Read tools.
+5. For every finding, construct a permalink.
-   - Find the files/functions mentioned or related to the bug.
+6. Write report to {REPORT_DIR}/issue-{number}.md
-   - Read them carefully and trace the logic.
-3. Determine one of three outcomes:
 
-OUTCOME A — CONFIRMED BUG (you found the problematic code):
+FINDING "ALREADY_FIXED" COMMITS:
-  Step 1: Post a comment on the issue. The comment MUST:
+- Use `git log --all --oneline -- {file}` to find recent changes to relevant files
-    - Start with exactly: [sisyphus-bot]
+- Use `git log --all --grep="fix" --grep="{keyword}" --all-match --oneline` to search commit messages
-    - Apologize sincerely for the inconvenience ("We're sorry you ran into this issue.")
+- Use `git blame {file}` to find who last changed the relevant lines
-    - Briefly acknowledge what the bug is
+- Use `git show {commit_sha}` to verify the fix
-    - Say "We've identified the root cause and will work on a fix."
+- Construct commit permalink: https://github.com/{REPO}/commit/{fix_commit_sha}
-    - Do NOT reveal internal implementation details unnecessarily
-  Step 2: Post the comment:
-    gh issue comment {number} --repo {REPO} --body "YOUR_COMMENT"
-  Step 3: Report back with:
-    ACTION: CONFIRMED_BUG
-    ROOT_CAUSE: [which file, which function, what goes wrong]
-    FIX_APPROACH: [how to fix it — be specific: "In {file}, line ~{N}, change X to Y because Z"]
-    SEVERITY: [LOW|MEDIUM|HIGH|CRITICAL]
-    AFFECTED_FILES: [list of files that need changes]
 
-OUTCOME B — NOT A BUG (user misunderstanding, provably correct behavior):
+REPORT FORMAT (write this as the file content):
-  ONLY choose this if you can RIGOROUSLY PROVE the behavior is correct.
-  Step 1: Post a comment. The comment MUST:
-    - Start with exactly: [sisyphus-bot]
-    - Be kind and empathetic — never condescending
-    - Explain clearly WHY the current behavior is correct
-    - Include specific code references or documentation links
-    - Offer a workaround or alternative if possible
-    - End with "Please let us know if you have further questions!"
-  Step 2: Post the comment:
-    gh issue comment {number} --repo {REPO} --body "YOUR_COMMENT"
-  Step 3: DO NOT close the issue. Let the user or maintainer decide.
-  Step 4: Report back with:
-    ACTION: NOT_A_BUG
-    EXPLANATION: [why this is correct behavior]
-    PROOF: [specific code reference proving it]
 
-OUTCOME C — UNCLEAR (can't determine from codebase alone):
+# Issue #{number}: {title}
-  Report back with:
+**Type:** Bug Report | **Author:** {author} | **Created:** {createdAt}
-    ACTION: NEEDS_INVESTIGATION
-    FINDINGS: [what you found so far]
-    BLOCKERS: [what's preventing you from determining the cause]
-    SUGGESTED_NEXT_STEPS: [what a human should look at]
 
-RULES:
+## Bug Summary
-- NEVER guess at root causes. Only report CONFIRMED_BUG if you found the exact problematic code.
+**Expected:** [what user expects]
-- NEVER close bug issues yourself. Only comment.
+**Actual:** [what actually happens]
-- For OUTCOME B (not a bug): you MUST have rigorous proof. If there's ANY doubt, choose OUTCOME C instead.
+**Reproduction:** [steps if provided]
-- The [sisyphus-bot] prefix is MANDATORY on every comment.
+
-- When apologizing, be genuine. The user took time to report this.
+## Verdict: [CONFIRMED_BUG | NOT_A_BUG | ALREADY_FIXED | UNCLEAR]
+
+## Analysis
+
+### Evidence
+[Each piece of evidence with permalink. No permalink = mark [UNVERIFIED]]
+
+### Root Cause (if CONFIRMED_BUG)
+[Which file, which function, what goes wrong]
+- Problematic code: [`{path}#L{N}`](permalink)
+
+### Why Not A Bug (if NOT_A_BUG)
+[Rigorous proof with permalinks that current behavior is correct]
+
+### Fix Details (if ALREADY_FIXED)
+- **Fixed in commit:** [`{short_sha}`](https://github.com/{REPO}/commit/{full_sha})
+- **Fixed date:** {date}
+- **What changed:** [description with diff permalink]
+- **Fixed by:** {author}
+
+### Blockers (if UNCLEAR)
+[What prevents determination, what to investigate next]
+
+## Severity: [LOW | MEDIUM | HIGH | CRITICAL]
+
+## Affected Files
+[List with permalinks]
+
+## Suggested Fix (if CONFIRMED_BUG)
+[Specific approach: "In {file}#L{N}, change X to Y because Z"]
+
+## Recommended Action
+[What maintainer should do]
+
+---
+CRITICAL: Claims without permalinks are worthless. If you cannot find evidence, say so explicitly rather than making unverified claims.
 ```
 
-</issue_bug_prompt>
-
 ---
 
-### SUBAGENT_ISSUE_FEATURE
+### ISSUE_FEATURE
-
-<issue_feature_prompt>
 
 ```
-You are a GitHub feature request analyzer for the repository {REPO}.
+You are analyzing feature request #{number} for {REPO}.
 
 ITEM:
 - Issue #{number}: {title}
@@ -235,38 +351,41 @@ ITEM:
 - Body: {body}
 - Comments: {comments_summary}
 
-YOUR JOB:
+TASK:
-1. Read the feature request.
+1. Understand the request.
-2. Search the codebase to check if this feature already exists (partially or fully).
+2. Search codebase for existing (partial/full) implementations.
-3. Assess feasibility and alignment with the project.
+3. Assess feasibility.
+4. Write report to {REPORT_DIR}/issue-{number}.md
 
-Report back with:
+REPORT FORMAT (write this as the file content):
-  ACTION: FEATURE_ASSESSED
-  ALREADY_EXISTS: [YES_FULLY | YES_PARTIALLY | NO]
-  IF_EXISTS: [where in the codebase, how to use it]
-  FEASIBILITY: [EASY | MODERATE | HARD | ARCHITECTURAL_CHANGE]
-  RELEVANT_FILES: [files that would need changes]
-  NOTES: [any observations about implementation approach]
 
-If the feature already fully exists:
+# Issue #{number}: {title}
-  Post a comment (prefix: [sisyphus-bot]) explaining how to use the existing feature with examples.
+**Type:** Feature Request | **Author:** {author} | **Created:** {createdAt}
-  gh issue comment {number} --repo {REPO} --body "YOUR_COMMENT"
 
-RULES:
+## Request Summary
-- Do NOT close feature requests.
+[What the user wants]
-- The [sisyphus-bot] prefix is MANDATORY on any comment.
+
+## Existing Implementation: [YES_FULLY | YES_PARTIALLY | NO]
+[If exists: where, with permalinks to the implementation]
+
+## Feasibility: [EASY | MODERATE | HARD | ARCHITECTURAL_CHANGE]
+
+## Relevant Files
+[With permalinks]
+
+## Implementation Notes
+[Approach, pitfalls, dependencies]
+
+## Recommended Action
+[What maintainer should do]
 ```
 
-</issue_feature_prompt>
-
 ---
 
-### SUBAGENT_ISSUE_OTHER
+### ISSUE_OTHER
-
-<issue_other_prompt>
 
 ```
-You are a GitHub issue analyzer for the repository {REPO}.
+You are analyzing issue #{number} for {REPO}.
 
 ITEM:
 - Issue #{number}: {title}
@@ -274,209 +393,195 @@ ITEM:
 - Body: {body}
 - Comments: {comments_summary}
 
-YOUR JOB:
+TASK: Assess and write report to {REPORT_DIR}/issue-{number}.md
-Quickly assess this issue and report:
-  ACTION: ASSESSED
-  TYPE_GUESS: [QUESTION | BUG | FEATURE | DISCUSSION | META | STALE]
-  SUMMARY: [1-2 sentence summary]
-  NEEDS_ATTENTION: [YES | NO]
-  SUGGESTED_LABEL: [if any]
 
-Do NOT post comments. Do NOT close. Just analyze and report.
+REPORT FORMAT (write this as the file content):
+
+# Issue #{number}: {title}
+**Type:** [QUESTION | BUG | FEATURE | DISCUSSION | META | STALE]
+**Author:** {author} | **Created:** {createdAt}
+
+## Summary
+[1-2 sentences]
+
+## Needs Attention: [YES | NO]
+## Suggested Label: [if any]
+## Recommended Action: [what maintainer should do]
 ```
 
-</issue_other_prompt>
-
 ---
 
-### SUBAGENT_PR_BUGFIX
+### PR_BUGFIX
-
-<pr_bugfix_prompt>
 
 ```
-You are a GitHub PR reviewer for the repository {REPO}.
+You are reviewing PR #{number} for {REPO}.
 
 ITEM:
 - PR #{number}: {title}
 - Author: {author}
-- Base: {baseRefName}
+- Base: {baseRefName} <- Head: {headRefName}
-- Head: {headRefName}
+- Draft: {isDraft} | Mergeable: {mergeable}
-- Draft: {isDraft}
+- Review: {reviewDecision} | CI: {statusCheckRollup_summary}
-- Mergeable: {mergeable}
-- Review Decision: {reviewDecision}
-- CI Status: {statusCheckRollup_summary}
 - Body: {body}
 
-YOUR JOB:
+TASK:
-1. Fetch PR details (DO NOT checkout the branch — read-only analysis):
+1. Fetch PR details (READ-ONLY): gh pr view {number} --repo {REPO} --json files,reviews,comments,statusCheckRollup,reviewDecision
-   gh pr view {number} --repo {REPO} --json files,reviews,comments,statusCheckRollup,reviewDecision
+2. Read diff: gh api repos/{REPO}/pulls/{number}/files
-2. Read the changed files list. For each changed file, use `gh api repos/{REPO}/pulls/{number}/files` to see the diff.
+3. Search codebase to verify fix correctness.
-3. Search the codebase to understand what the PR is fixing and whether the fix is correct.
+4. Write report to {REPORT_DIR}/pr-{number}.md
-4. Evaluate merge safety:
 
-MERGE CONDITIONS (ALL must be true for auto-merge):
+REPORT FORMAT (write this as the file content):
-  a. CI status checks: ALL passing (no failures, no pending)
-  b. Review decision: APPROVED
-  c. The fix is clearly correct — addresses an obvious, unambiguous bug
-  d. No risky side effects (no architectural changes, no breaking changes)
-  e. Not a draft PR
-  f. Mergeable state is clean (no conflicts)
 
-IF ALL MERGE CONDITIONS MET:
+# PR #{number}: {title}
-  Step 1: Merge the PR:
+**Type:** Bugfix | **Author:** {author}
-    gh pr merge {number} --repo {REPO} --squash --auto
+**Base:** {baseRefName} <- {headRefName} | **Draft:** {isDraft}
-  Step 2: Report back with:
-    ACTION: MERGED
-    FIX_SUMMARY: [what bug was fixed and how]
-    FILES_CHANGED: [list of files]
-    RISK: NONE
 
-IF ANY CONDITION NOT MET:
+## Fix Summary
-  Report back with:
+[What bug, how fixed - with permalinks to changed code]
-    ACTION: NEEDS_HUMAN_DECISION
-    FIX_SUMMARY: [what the PR does]
-    WHAT_IT_FIXES: [the bug or issue it addresses]
-    CI_STATUS: [PASS | FAIL | PENDING — list any failures]
-    REVIEW_STATUS: [APPROVED | CHANGES_REQUESTED | PENDING | NONE]
-    MISSING: [what's preventing auto-merge — be specific]
-    RISK_ASSESSMENT: [what could go wrong]
-    AMBIGUOUS_PARTS: [anything that needs human judgment]
-    RECOMMENDED_ACTION: [what the maintainer should do]
 
-ABSOLUTE RULES:
+## Code Review
-- NEVER run `git checkout`, `git fetch`, `git pull`, or `git switch`. READ-ONLY via gh CLI and API.
+
-- NEVER checkout the PR branch. NEVER. Use `gh api` and `gh pr view` only.
+### Correctness
-- Only merge if you are 100% certain ALL conditions are met. When in doubt, report instead.
+[Is fix correct? Root cause addressed? Evidence with permalinks]
-- The [sisyphus-bot] prefix is MANDATORY on any comment you post.
+
+### Side Effects
+[Risky changes, breaking changes - with permalinks if any]
+
+### Code Quality
+[Style, patterns, test coverage]
+
+## Merge Readiness
+
+| Check | Status |
+|-------|--------|
+| CI | [PASS / FAIL / PENDING] |
+| Review | [APPROVED / CHANGES_REQUESTED / PENDING / NONE] |
+| Mergeable | [YES / NO / CONFLICTED] |
+| Draft | [YES / NO] |
+| Correctness | [VERIFIED / CONCERNS / UNCLEAR] |
+| Risk | [NONE / LOW / MEDIUM / HIGH] |
+
+## Files Changed
+[List with brief descriptions]
+
+## Recommended Action: [MERGE | REQUEST_CHANGES | NEEDS_REVIEW | WAIT]
+[Reasoning with evidence]
+
+---
+NEVER merge. NEVER comment. NEVER review. Write to file ONLY.
 ```
 
-</pr_bugfix_prompt>
-
 ---
 
-### SUBAGENT_PR_OTHER
+### PR_OTHER
-
-<pr_other_prompt>
 
 ```
-You are a GitHub PR reviewer for the repository {REPO}.
+You are reviewing PR #{number} for {REPO}.
 
 ITEM:
 - PR #{number}: {title}
 - Author: {author}
-- Base: {baseRefName}
+- Base: {baseRefName} <- Head: {headRefName}
-- Head: {headRefName}
+- Draft: {isDraft} | Mergeable: {mergeable}
-- Draft: {isDraft}
+- Review: {reviewDecision} | CI: {statusCheckRollup_summary}
-- Mergeable: {mergeable}
-- Review Decision: {reviewDecision}
-- CI Status: {statusCheckRollup_summary}
 - Body: {body}
 
-YOUR JOB:
+TASK:
-1. Fetch PR details (READ-ONLY — no checkout):
+1. Fetch PR details (READ-ONLY): gh pr view {number} --repo {REPO} --json files,reviews,comments,statusCheckRollup,reviewDecision
-   gh pr view {number} --repo {REPO} --json files,reviews,comments,statusCheckRollup,reviewDecision
+2. Read diff: gh api repos/{REPO}/pulls/{number}/files
-2. Read the changed files via `gh api repos/{REPO}/pulls/{number}/files`.
+3. Write report to {REPORT_DIR}/pr-{number}.md
-3. Assess the PR and report:
 
-  ACTION: PR_ASSESSED
+REPORT FORMAT (write this as the file content):
-  TYPE: [FEATURE | REFACTOR | DOCS | CHORE | TEST | OTHER]
-  SUMMARY: [what this PR does in 2-3 sentences]
-  CI_STATUS: [PASS | FAIL | PENDING]
-  REVIEW_STATUS: [APPROVED | CHANGES_REQUESTED | PENDING | NONE]
-  FILES_CHANGED: [count and key files]
-  RISK_LEVEL: [LOW | MEDIUM | HIGH]
-  ALIGNMENT: [does this fit the project direction? YES | NO | UNCLEAR]
-  BLOCKERS: [anything preventing merge]
-  RECOMMENDED_ACTION: [MERGE | REQUEST_CHANGES | NEEDS_REVIEW | CLOSE | WAIT]
-  NOTES: [any observations for the maintainer]
 
-ABSOLUTE RULES:
+# PR #{number}: {title}
-- NEVER run `git checkout`, `git fetch`, `git pull`, or `git switch`. READ-ONLY.
+**Type:** [FEATURE | REFACTOR | DOCS | CHORE | TEST | OTHER]
-- NEVER checkout the PR branch. Use `gh api` and `gh pr view` only.
+**Author:** {author}
-- Do NOT merge non-bugfix PRs automatically. Report only.
+**Base:** {baseRefName} <- {headRefName} | **Draft:** {isDraft}
+
+## Summary
+[2-3 sentences with permalinks to key changes]
+
+## Status
+
+| Check | Status |
+|-------|--------|
+| CI | [PASS / FAIL / PENDING] |
+| Review | [APPROVED / CHANGES_REQUESTED / PENDING / NONE] |
+| Mergeable | [YES / NO / CONFLICTED] |
+| Risk | [LOW / MEDIUM / HIGH] |
+| Alignment | [YES / NO / UNCLEAR] |
+
+## Files Changed
+[Count and key files]
+
+## Blockers
+[If any]
+
+## Recommended Action: [MERGE | REQUEST_CHANGES | NEEDS_REVIEW | CLOSE | WAIT]
+[Reasoning]
+
+---
+NEVER merge. NEVER comment. NEVER review. Write to file ONLY.
 ```
 
-</pr_other_prompt>
+---
+
+## Phase 4: Collect & Update
+
+Poll `background_output()` per task. As each completes:
+1. Parse report.
+2. `task_update(id=task_id, status="completed", description=REPORT_SUMMARY)`
+3. Stream to user immediately.
 
 ---
 
-## PHASE 4: COLLECT RESULTS & UPDATE TASKS
+## Phase 5: Final Summary
 
-<collection>
+Write to `{REPORT_DIR}/SUMMARY.md` AND display to user:
-Poll `background_output()` for each spawned task. As each completes:
-
-1. Parse the subagent's report.
-2. Update the corresponding TaskCreate entry:
-   - `TaskUpdate(id=task_id, status="completed", description=FULL_REPORT_TEXT)`
-3. Stream the result to the user immediately — do not wait for all to finish.
-
-Track counters:
-- issues_answered (commented + closed)
-- bugs_confirmed
-- bugs_not_a_bug
-- prs_merged
-- prs_needs_decision
-- features_assessed
-</collection>
-
----
-
-## PHASE 5: FINAL SUMMARY
-
-After all background tasks complete, produce a summary:
 
 ```markdown
-# GitHub Triage Report — {REPO}
+# GitHub Triage Report - {REPO}
 
-**Date:** {date}
+**Date:** {date} | **Commit:** {COMMIT_SHA}
 **Items Processed:** {total}
+**Report Directory:** {REPORT_DIR}
 
 ## Issues ({issue_count})
-| Action | Count |
+| Category | Count |
-|--------|-------|
+|----------|-------|
-| Answered & Closed | {issues_answered} |
+| Bug Confirmed | {n} |
-| Bug Confirmed | {bugs_confirmed} |
+| Bug Already Fixed | {n} |
-| Not A Bug (explained) | {bugs_not_a_bug} |
+| Not A Bug | {n} |
-| Feature Assessed | {features_assessed} |
+| Needs Investigation | {n} |
-| Needs Manual Attention | {needs_manual} |
+| Question Analyzed | {n} |
+| Feature Assessed | {n} |
+| Other | {n} |
 
 ## PRs ({pr_count})
-| Action | Count |
+| Category | Count |
-|--------|-------|
+|----------|-------|
-| Auto-Merged (safe bugfix) | {prs_merged} |
+| Bugfix Reviewed | {n} |
-| Needs Human Decision | {prs_needs_decision} |
+| Other PR Reviewed | {n} |
-| Assessed (non-bugfix) | {prs_assessed} |
 
-## Items Requiring Your Attention
+## Items Requiring Attention
-[List each item that needs human decision with its report summary]
+[Each item: number, title, verdict, 1-line summary, link to report file]
+
+## Report Files
+[All generated files with paths]
 ```
 
 ---
 
-## ANTI-PATTERNS
+## Anti-Patterns
 
 | Violation | Severity |
 |-----------|----------|
-| Using any category other than `free` | CRITICAL |
+| ANY GitHub mutation (comment/close/merge/review/label/edit) | **CRITICAL** |
+| Claim without permalink | **CRITICAL** |
+| Using category other than `quick` | CRITICAL |
 | Batching multiple items into one task | CRITICAL |
-| Using `run_in_background=false` | CRITICAL |
+| `run_in_background=false` | CRITICAL |
-| Subagent running `git checkout` on a PR branch | CRITICAL |
+| `git checkout` on PR branch | CRITICAL |
-| Posting comment without `[sisyphus-bot]` prefix | CRITICAL |
+| Guessing without codebase evidence | HIGH |
-| Merging a PR that doesn't meet ALL 6 conditions | CRITICAL |
+| Not writing report to `{REPORT_DIR}` | HIGH |
-| Closing a bug issue (only comment, never close bugs) | HIGH |
+| Using branch name instead of commit SHA in permalink | HIGH |
-| Guessing at answers without codebase evidence | HIGH |
-| Not recording results via TaskCreate/TaskUpdate | HIGH |
-
----
-
-## QUICK START
-
-When invoked:
-
-1. `TaskCreate` for the overall triage job
-2. Fetch all open issues + PRs via gh CLI (paginate if needed)
-3. Classify each item (ISSUE_QUESTION, ISSUE_BUG, ISSUE_FEATURE, PR_BUGFIX, etc.)
-4. For EACH item: `TaskCreate` + `task(category="free", run_in_background=true, load_skills=[], prompt=...)`
-5. Poll `background_output()` — stream results as they arrive
-6. `TaskUpdate` each task with the subagent's findings
-7. Produce final summary report

.opencode/skills/pre-publish-review/SKILL.md

@@ -0,0 +1,407 @@
+---
+name: pre-publish-review
+description: "Nuclear-grade 16-agent pre-publish release gate. Runs /get-unpublished-changes to detect all changes since last npm release, spawns up to 10 ultrabrain agents for deep per-change analysis, invokes /review-work (5 agents) for holistic review, and 1 oracle for overall release synthesis. Use before EVERY npm publish. Triggers: 'pre-publish review', 'review before publish', 'release review', 'pre-release review', 'ready to publish?', 'can I publish?', 'pre-publish', 'safe to publish', 'publishing review', 'pre-publish check'."
+---
+
+# Pre-Publish Review — 16-Agent Release Gate
+
+Three-layer review before publishing to npm. Every layer covers a different angle — together they catch what no single reviewer could.
+
+| Layer | Agents | Type | What They Check |
+|-------|--------|------|-----------------|
+| Per-Change Deep Dive | up to 10 | ultrabrain | Each logical change group individually — correctness, edge cases, pattern adherence |
+| Holistic Review | 5 | review-work | Goal compliance, QA execution, code quality, security, context mining across full changeset |
+| Release Synthesis | 1 | oracle | Overall release readiness, version bump, breaking changes, deployment risk |
+
+---
+
+## Phase 0: Detect Unpublished Changes
+
+Run `/get-unpublished-changes` FIRST. This is the single source of truth for what changed.
+
+```
+skill(name="get-unpublished-changes")
+```
+
+This command automatically:
+- Detects published npm version vs local version
+- Lists all commits since last release
+- Reads actual diffs (not just commit messages) to describe REAL changes
+- Groups changes by type (feat/fix/refactor/docs) with scope
+- Identifies breaking changes
+- Recommends version bump (patch/minor/major)
+
+**Save the full output** — it feeds directly into Phase 1 grouping and all agent prompts.
+
+Then capture raw data needed by agent prompts:
+
+```bash
+# Extract versions (already in /get-unpublished-changes output)
+PUBLISHED=$(npm view oh-my-opencode version 2>/dev/null || echo "not published")
+LOCAL=$(node -p "require('./package.json').version" 2>/dev/null || echo "unknown")
+
+# Raw data for agents (diffs, file lists)
+COMMITS=$(git log "v${PUBLISHED}"..HEAD --oneline 2>/dev/null || echo "no commits")
+COMMIT_COUNT=$(echo "$COMMITS" | wc -l | tr -d ' ')
+DIFF_STAT=$(git diff "v${PUBLISHED}"..HEAD --stat 2>/dev/null || echo "no diff")
+CHANGED_FILES=$(git diff --name-only "v${PUBLISHED}"..HEAD 2>/dev/null || echo "none")
+FILE_COUNT=$(echo "$CHANGED_FILES" | wc -l | tr -d ' ')
+```
+
+If `PUBLISHED` is "not published", this is a first release — use the full git history instead.
+---
+
+## Phase 1: Parse Changes into Groups
+
+Use the `/get-unpublished-changes` output as the starting point — it already groups by scope and type.
+
+**Grouping strategy:**
+1. Start from the `/get-unpublished-changes` analysis which already categorizes by feat/fix/refactor/docs with scope
+2. Further split by **module/area** — changes touching the same module or feature area belong together
+3. Target **up to 10 groups**. If fewer than 10 commits, each commit is its own group. If more than 10 logical areas, merge the smallest groups.
+4. For each group, extract:
+   - **Group name**: Short descriptive label (e.g., "agent-model-resolution", "hook-system-refactor")
+   - **Commits**: List of commit hashes and messages
+   - **Files**: Changed files in this group
+   - **Diff**: The relevant portion of the full diff (`git diff v${PUBLISHED}..HEAD -- {group files}`)
+
+---
+
+## Phase 2: Spawn All Agents
+
+Launch ALL agents in a single turn. Every agent uses `run_in_background=true`. No sequential launches.
+
+### Layer 1: Ultrabrain Per-Change Analysis (up to 10)
+
+For each change group, spawn one ultrabrain agent. Each gets only its portion of the diff — not the full changeset.
+
+```
+task(
+  category="ultrabrain",
+  run_in_background=true,
+  load_skills=[],
+  description="Deep analysis: {GROUP_NAME}",
+  prompt="""
+<review_type>PER-CHANGE DEEP ANALYSIS</review_type>
+<change_group>{GROUP_NAME}</change_group>
+
+<project>oh-my-opencode (npm package)</project>
+<published_version>{PUBLISHED}</published_version>
+<target_version>{LOCAL}</target_version>
+
+<commits>
+{GROUP_COMMITS — hash and message for each commit in this group}
+</commits>
+
+<changed_files>
+{GROUP_FILES — files changed in this group}
+</changed_files>
+
+<diff>
+{GROUP_DIFF — only the diff for this group's files}
+</diff>
+
+<file_contents>
+{Read and include full content of each changed file in this group}
+</file_contents>
+
+You are reviewing a specific subset of changes heading into an npm release. Focus exclusively on THIS change group. Other groups are reviewed by parallel agents.
+
+ANALYSIS CHECKLIST:
+
+1. **Intent Clarity**: What is this change trying to do? Is the intent clear from the code and commit messages? If you have to guess, that's a finding.
+
+2. **Correctness**: Trace through the logic for 3+ scenarios. Does the code actually do what it claims? Off-by-one errors, null handling, async edge cases, resource cleanup.
+
+3. **Breaking Changes**: Does this change alter any public API, config format, CLI behavior, or hook contract? If yes, is it backward compatible? Would existing users be surprised?
+
+4. **Pattern Adherence**: Does the new code follow the established patterns visible in the existing file contents? New patterns where old ones exist = finding.
+
+5. **Edge Cases**: What inputs or conditions would break this? Empty arrays, undefined values, concurrent calls, very large inputs, missing config fields.
+
+6. **Error Handling**: Are errors properly caught and propagated? No empty catch blocks? No swallowed promises?
+
+7. **Type Safety**: Any `as any`, `@ts-ignore`, `@ts-expect-error`? Loose typing where strict is possible?
+
+8. **Test Coverage**: Are the behavioral changes covered by tests? Are the tests meaningful or just coverage padding?
+
+9. **Side Effects**: Could this change break something in a different module? Check imports and exports — who depends on what changed?
+
+10. **Release Risk**: On a scale of SAFE / CAUTION / RISKY — how confident are you this change won't cause issues in production?
+
+OUTPUT FORMAT:
+<group_name>{GROUP_NAME}</group_name>
+<verdict>PASS or FAIL</verdict>
+<risk>SAFE / CAUTION / RISKY</risk>
+<summary>2-3 sentence assessment of this change group</summary>
+<has_breaking_changes>YES or NO</has_breaking_changes>
+<breaking_change_details>If YES, describe what breaks and for whom</breaking_change_details>
+<findings>
+  For each finding:
+  - [CRITICAL/MAJOR/MINOR] Category: Description
+  - File: path (line range)
+  - Evidence: specific code reference
+  - Suggestion: how to fix
+</findings>
+<blocking_issues>Issues that MUST be fixed before publish. Empty if PASS.</blocking_issues>
+""")
+```
+
+### Layer 2: Holistic Review via /review-work (5 agents)
+
+Spawn a sub-agent that loads the `/review-work` skill. The review-work skill internally launches 5 parallel agents: Oracle (goal verification), unspecified-high (QA execution), Oracle (code quality), Oracle (security), unspecified-high (context mining). All 5 must pass for the review to pass.
+
+```
+task(
+  category="unspecified-high",
+  run_in_background=true,
+  load_skills=["review-work"],
+  description="Run /review-work on all unpublished changes",
+  prompt="""
+Run /review-work on the unpublished changes between v{PUBLISHED} and HEAD.
+
+GOAL: Review all changes heading into npm publish of oh-my-opencode. These changes span {COMMIT_COUNT} commits across {FILE_COUNT} files.
+
+CONSTRAINTS:
+- This is a plugin published to npm — public API stability matters
+- TypeScript strict mode, Bun runtime
+- No `as any`, `@ts-ignore`, `@ts-expect-error`
+- Factory pattern (createXXX) for tools, hooks, agents
+- kebab-case files, barrel exports, no catch-all files
+
+BACKGROUND: Pre-publish review of oh-my-opencode, an OpenCode plugin with 1268 TypeScript files, 160k LOC. Changes since v{PUBLISHED} are about to be published.
+
+The diff base is: git diff v{PUBLISHED}..HEAD
+
+Follow the /review-work skill flow exactly — launch all 5 review agents and collect results. Do NOT skip any of the 5 agents.
+""")
+```
+
+### Layer 3: Oracle Release Synthesis (1 agent)
+
+The oracle gets the full picture — all commits, full diff stat, and changed file list. It provides the final release readiness assessment.
+
+```
+task(
+  subagent_type="oracle",
+  run_in_background=true,
+  load_skills=[],
+  description="Oracle: overall release synthesis and version bump recommendation",
+  prompt="""
+<review_type>RELEASE SYNTHESIS — OVERALL ASSESSMENT</review_type>
+
+<project>oh-my-opencode (npm package)</project>
+<published_version>{PUBLISHED}</published_version>
+<local_version>{LOCAL}</local_version>
+
+<all_commits>
+{ALL COMMITS since published version — hash, message, author, date}
+</all_commits>
+
+<diff_stat>
+{DIFF_STAT — files changed, insertions, deletions}
+</diff_stat>
+
+<changed_files>
+{CHANGED_FILES — full list of modified file paths}
+</changed_files>
+
+<full_diff>
+{FULL_DIFF — the complete git diff between published version and HEAD}
+</full_diff>
+
+<file_contents>
+{Read and include full content of KEY changed files — focus on public API surfaces, config schemas, agent definitions, hook registrations, tool registrations}
+</file_contents>
+
+You are the final gate before an npm publish. 10 ultrabrain agents are reviewing individual changes and 5 review-work agents are doing holistic review. Your job is the bird's-eye view that those focused reviews might miss.
+
+SYNTHESIS CHECKLIST:
+
+1. **Release Coherence**: Do these changes tell a coherent story? Or is this a grab-bag of unrelated changes that should be split into multiple releases?
+
+2. **Version Bump**: Based on semver:
+   - PATCH: Bug fixes only, no behavior changes
+   - MINOR: New features, backward-compatible changes
+   - MAJOR: Breaking changes to public API, config format, or behavior
+   Recommend the correct bump with specific justification.
+
+3. **Breaking Changes Audit**: Exhaustively list every change that could break existing users. Check:
+   - Config schema changes (new required fields, removed fields, renamed fields)
+   - Agent behavior changes (different prompts, different model routing)
+   - Hook contract changes (new parameters, removed hooks, renamed hooks)
+   - Tool interface changes (new required params, different return types)
+   - CLI changes (new commands, changed flags, different output)
+   - Skill format changes (SKILL.md schema changes)
+
+4. **Migration Requirements**: If there are breaking changes, what migration steps do users need? Is there auto-migration in place?
+
+5. **Dependency Changes**: New dependencies added? Dependencies removed? Version bumps? Any supply chain risk?
+
+6. **Changelog Draft**: Write a draft changelog entry grouped by:
+   - feat: New features
+   - fix: Bug fixes
+   - refactor: Internal changes (no user impact)
+   - breaking: Breaking changes with migration instructions
+   - docs: Documentation changes
+
+7. **Deployment Risk Assessment**:
+   - SAFE: Routine changes, well-tested, low risk
+   - CAUTION: Significant changes but manageable risk
+   - RISKY: Large surface area changes, insufficient testing, or breaking changes without migration
+   - BLOCK: Critical issues found, do NOT publish
+
+8. **Post-Publish Monitoring**: What should be monitored after publish? Error rates, specific features, user feedback channels.
+
+OUTPUT FORMAT:
+<verdict>SAFE / CAUTION / RISKY / BLOCK</verdict>
+<recommended_version_bump>PATCH / MINOR / MAJOR</recommended_version_bump>
+<version_bump_justification>Why this bump level</version_bump_justification>
+<release_coherence>Assessment of whether changes belong in one release</release_coherence>
+<breaking_changes>
+  Exhaustive list, or "None" if none.
+  For each:
+  - What changed
+  - Who is affected
+  - Migration steps
+</breaking_changes>
+<changelog_draft>
+  Ready-to-use changelog entry
+</changelog_draft>
+<deployment_risk>
+  Overall risk assessment with specific concerns
+</deployment_risk>
+<monitoring_recommendations>
+  What to watch after publish
+</monitoring_recommendations>
+<blocking_issues>Issues that MUST be fixed before publish. Empty if SAFE.</blocking_issues>
+""")
+```
+
+---
+
+## Phase 3: Collect Results
+
+As agents complete (system notifications), collect via `background_output(task_id="...")`.
+
+Track completion in a table:
+
+| # | Agent | Type | Status | Verdict |
+|---|-------|------|--------|---------|
+| 1-10 | Ultrabrain: {group_name} | ultrabrain | pending | — |
+| 11 | Review-Work Coordinator | unspecified-high | pending | — |
+| 12 | Release Synthesis Oracle | oracle | pending | — |
+
+Do NOT deliver the final report until ALL agents have completed.
+
+---
+
+## Phase 4: Final Verdict
+
+<verdict_logic>
+
+**BLOCK** if:
+- Oracle verdict is BLOCK
+- Any ultrabrain found CRITICAL blocking issues
+- Review-work failed on any MAIN agent
+
+**RISKY** if:
+- Oracle verdict is RISKY
+- Multiple ultrabrains returned CAUTION or FAIL
+- Review-work passed but with significant findings
+
+**CAUTION** if:
+- Oracle verdict is CAUTION
+- A few ultrabrains flagged minor issues
+- Review-work passed cleanly
+
+**SAFE** if:
+- Oracle verdict is SAFE
+- All ultrabrains passed
+- Review-work passed
+
+</verdict_logic>
+
+Compile the final report:
+
+```markdown
+# Pre-Publish Review — oh-my-opencode
+
+## Release: v{PUBLISHED} -> v{LOCAL}
+**Commits:** {COMMIT_COUNT} | **Files Changed:** {FILE_COUNT} | **Agents:** {AGENT_COUNT}
+
+---
+
+## Overall Verdict: SAFE / CAUTION / RISKY / BLOCK
+
+## Recommended Version Bump: PATCH / MINOR / MAJOR
+{Justification from Oracle}
+
+---
+
+## Per-Change Analysis (Ultrabrains)
+
+| # | Change Group | Verdict | Risk | Breaking? | Blocking Issues |
+|---|-------------|---------|------|-----------|-----------------|
+| 1 | {name} | PASS/FAIL | SAFE/CAUTION/RISKY | YES/NO | {count or "none"} |
+| ... | ... | ... | ... | ... | ... |
+
+### Blocking Issues from Per-Change Analysis
+{Aggregated from all ultrabrains — deduplicated}
+
+---
+
+## Holistic Review (Review-Work)
+
+| # | Review Area | Verdict | Confidence |
+|---|------------|---------|------------|
+| 1 | Goal & Constraint Verification | PASS/FAIL | HIGH/MED/LOW |
+| 2 | QA Execution | PASS/FAIL | HIGH/MED/LOW |
+| 3 | Code Quality | PASS/FAIL | HIGH/MED/LOW |
+| 4 | Security | PASS/FAIL | Severity |
+| 5 | Context Mining | PASS/FAIL | HIGH/MED/LOW |
+
+### Blocking Issues from Holistic Review
+{Aggregated from review-work}
+
+---
+
+## Release Synthesis (Oracle)
+
+### Breaking Changes
+{From Oracle — exhaustive list or "None"}
+
+### Changelog Draft
+{From Oracle — ready to use}
+
+### Deployment Risk
+{From Oracle — specific concerns}
+
+### Post-Publish Monitoring
+{From Oracle — what to watch}
+
+---
+
+## All Blocking Issues (Prioritized)
+{Deduplicated, merged from all three layers, ordered by severity}
+
+## Recommendations
+{If BLOCK/RISKY: exactly what to fix, in priority order}
+{If CAUTION: suggestions worth considering before publish}
+{If SAFE: non-blocking improvements for future}
+```
+
+---
+
+## Anti-Patterns
+
+| Violation | Severity |
+|-----------|----------|
+| Publishing without waiting for all agents | **CRITICAL** |
+| Spawning ultrabrains sequentially instead of in parallel | CRITICAL |
+| Using `run_in_background=false` for any agent | CRITICAL |
+| Skipping the Oracle synthesis | HIGH |
+| Not reading file contents for Oracle (it cannot read files) | HIGH |
+| Grouping all changes into 1-2 ultrabrains instead of distributing | HIGH |
+| Delivering verdict before all agents complete | HIGH |
+| Not including diff in ultrabrain prompts | MAJOR |

.opencode/skills/work-with-pr-workspace/evals/evals.json

@@ -0,0 +1,76 @@
+{
+  "skill_name": "work-with-pr",
+  "evals": [
+    {
+      "id": 1,
+      "prompt": "I need to add a `max_background_agents` config option to oh-my-opencode that limits how many background agents can run simultaneously. It should be in the plugin config schema with a default of 5. Add validation and make sure the background manager respects it. Create a PR for this.",
+      "expected_output": "Agent creates worktree, implements config option with schema validation, adds tests, creates PR, iterates through verification gates until merged",
+      "files": [],
+      "assertions": [
+        {"id": "worktree-isolation", "text": "Plan uses git worktree in a sibling directory (not main working directory)"},
+        {"id": "branch-from-dev", "text": "Branch is created from origin/dev (not master/main)"},
+        {"id": "atomic-commits", "text": "Plan specifies multiple atomic commits for multi-file changes"},
+        {"id": "local-validation", "text": "Runs bun run typecheck, bun test, and bun run build before pushing"},
+        {"id": "pr-targets-dev", "text": "PR is created targeting dev branch (not master)"},
+        {"id": "three-gates", "text": "Verification loop includes all 3 gates: CI, review-work, and Cubic"},
+        {"id": "gate-ordering", "text": "Gates are checked in order: CI first, then review-work, then Cubic"},
+        {"id": "cubic-check-method", "text": "Cubic check uses gh api to check cubic-dev-ai[bot] reviews for 'No issues found'"},
+        {"id": "worktree-cleanup", "text": "Plan includes worktree cleanup after merge"},
+        {"id": "real-file-references", "text": "Code changes reference actual files in the codebase (config schema, background manager)"}
+      ]
+    },
+    {
+      "id": 2,
+      "prompt": "The atlas hook has a bug where it crashes when boulder.json is missing the worktree_path field. Fix it and land the fix as a PR. Make sure CI passes.",
+      "expected_output": "Agent creates worktree for the fix branch, adds null check and test for missing worktree_path, creates PR, iterates verification loop",
+      "files": [],
+      "assertions": [
+        {"id": "worktree-isolation", "text": "Plan uses git worktree in a sibling directory"},
+        {"id": "minimal-fix", "text": "Fix is minimal — adds null check, doesn't refactor unrelated code"},
+        {"id": "test-added", "text": "Test case added for the missing worktree_path scenario"},
+        {"id": "three-gates", "text": "Verification loop includes all 3 gates: CI, review-work, Cubic"},
+        {"id": "real-atlas-files", "text": "References actual atlas hook files in src/hooks/atlas/"},
+        {"id": "fix-branch-naming", "text": "Branch name follows fix/ prefix convention"}
+      ]
+    },
+    {
+      "id": 3,
+      "prompt": "Refactor src/tools/delegate-task/constants.ts to split DEFAULT_CATEGORIES and CATEGORY_MODEL_REQUIREMENTS into separate files. Keep backward compatibility with the barrel export. Make a PR.",
+      "expected_output": "Agent creates worktree, splits file with atomic commits, ensures imports still work via barrel, creates PR, runs through all gates",
+      "files": [],
+      "assertions": [
+        {"id": "worktree-isolation", "text": "Plan uses git worktree in a sibling directory"},
+        {"id": "multiple-atomic-commits", "text": "Uses 2+ commits for the multi-file refactor"},
+        {"id": "barrel-export", "text": "Maintains backward compatibility via barrel re-export in constants.ts or index.ts"},
+        {"id": "three-gates", "text": "Verification loop includes all 3 gates"},
+        {"id": "real-constants-file", "text": "References actual src/tools/delegate-task/constants.ts file and its exports"}
+      ]
+    },
+    {
+      "id": 4,
+      "prompt": "implement issue #100 - we need to add a new built-in MCP for arxiv paper search. just the basic search endpoint, nothing fancy. pr it",
+      "expected_output": "Agent creates worktree, implements arxiv MCP following existing MCP patterns (websearch, context7, grep_app), creates PR with proper template, verification loop runs",
+      "files": [],
+      "assertions": [
+        {"id": "worktree-isolation", "text": "Plan uses git worktree in a sibling directory"},
+        {"id": "follows-mcp-pattern", "text": "New MCP follows existing pattern from src/mcp/ (websearch, context7, grep_app)"},
+        {"id": "three-gates", "text": "Verification loop includes all 3 gates"},
+        {"id": "pr-targets-dev", "text": "PR targets dev branch"},
+        {"id": "local-validation", "text": "Runs local checks before pushing"}
+      ]
+    },
+    {
+      "id": 5,
+      "prompt": "The comment-checker hook is too aggressive - it's flagging legitimate comments that happen to contain 'Note:' as AI slop. Relax the regex pattern and add test cases for the false positives. Work on a separate branch and make a PR.",
+      "expected_output": "Agent creates worktree, fixes regex, adds specific test cases for false positive scenarios, creates PR, all three gates pass",
+      "files": [],
+      "assertions": [
+        {"id": "worktree-isolation", "text": "Plan uses git worktree in a sibling directory"},
+        {"id": "real-comment-checker-files", "text": "References actual comment-checker hook files in the codebase"},
+        {"id": "regression-tests", "text": "Adds test cases specifically for 'Note:' false positive scenarios"},
+        {"id": "three-gates", "text": "Verification loop includes all 3 gates"},
+        {"id": "minimal-change", "text": "Only modifies regex and adds tests — no unrelated changes"}
+      ]
+    }
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/benchmark.json

@@ -0,0 +1,138 @@
+{
+  "skill_name": "work-with-pr",
+  "iteration": 1,
+  "summary": {
+    "with_skill": {
+      "pass_rate": 0.968,
+      "mean_duration_seconds": 340.2,
+      "stddev_duration_seconds": 169.3
+    },
+    "without_skill": {
+      "pass_rate": 0.516,
+      "mean_duration_seconds": 303.0,
+      "stddev_duration_seconds": 77.8
+    },
+    "delta": {
+      "pass_rate": 0.452,
+      "mean_duration_seconds": 37.2,
+      "stddev_duration_seconds": 91.5
+    }
+  },
+  "evals": [
+    {
+      "eval_name": "happy-path-feature-config-option",
+      "with_skill": {
+        "pass_rate": 1.0,
+        "passed": 10,
+        "total": 10,
+        "duration_seconds": 292,
+        "failed_assertions": []
+      },
+      "without_skill": {
+        "pass_rate": 0.4,
+        "passed": 4,
+        "total": 10,
+        "duration_seconds": 365,
+        "failed_assertions": [
+          {"assertion": "Plan uses git worktree in a sibling directory", "reason": "Uses git checkout -b, no worktree isolation"},
+          {"assertion": "Plan specifies multiple atomic commits for multi-file changes", "reason": "Steps listed sequentially but no atomic commit strategy mentioned"},
+          {"assertion": "Verification loop includes all 3 gates: CI, review-work, and Cubic", "reason": "Only mentions CI pipeline in step 6. No review-work or Cubic."},
+          {"assertion": "Gates are checked in order: CI first, then review-work, then Cubic", "reason": "No gate ordering - only CI mentioned"},
+          {"assertion": "Cubic check uses gh api to check cubic-dev-ai[bot] reviews", "reason": "No mention of Cubic at all"},
+          {"assertion": "Plan includes worktree cleanup after merge", "reason": "No worktree used, no cleanup needed"}
+        ]
+      }
+    },
+    {
+      "eval_name": "bugfix-atlas-null-check",
+      "with_skill": {
+        "pass_rate": 1.0,
+        "passed": 6,
+        "total": 6,
+        "duration_seconds": 506,
+        "failed_assertions": []
+      },
+      "without_skill": {
+        "pass_rate": 0.667,
+        "passed": 4,
+        "total": 6,
+        "duration_seconds": 325,
+        "failed_assertions": [
+          {"assertion": "Plan uses git worktree in a sibling directory", "reason": "No worktree. Steps go directly to creating branch and modifying files."},
+          {"assertion": "Verification loop includes all 3 gates", "reason": "Only mentions CI pipeline (step 5). No review-work or Cubic."}
+        ]
+      }
+    },
+    {
+      "eval_name": "refactor-split-constants",
+      "with_skill": {
+        "pass_rate": 1.0,
+        "passed": 5,
+        "total": 5,
+        "duration_seconds": 181,
+        "failed_assertions": []
+      },
+      "without_skill": {
+        "pass_rate": 0.4,
+        "passed": 2,
+        "total": 5,
+        "duration_seconds": 229,
+        "failed_assertions": [
+          {"assertion": "Plan uses git worktree in a sibling directory", "reason": "git checkout -b only, no worktree"},
+          {"assertion": "Uses 2+ commits for the multi-file refactor", "reason": "Single atomic commit: 'refactor: split delegate-task constants and category model requirements'"},
+          {"assertion": "Verification loop includes all 3 gates", "reason": "Only mentions typecheck/test/build. No review-work or Cubic."}
+        ]
+      }
+    },
+    {
+      "eval_name": "new-mcp-arxiv-casual",
+      "with_skill": {
+        "pass_rate": 1.0,
+        "passed": 5,
+        "total": 5,
+        "duration_seconds": 152,
+        "failed_assertions": []
+      },
+      "without_skill": {
+        "pass_rate": 0.6,
+        "passed": 3,
+        "total": 5,
+        "duration_seconds": 197,
+        "failed_assertions": [
+          {"assertion": "Verification loop includes all 3 gates", "reason": "Only mentions bun test/typecheck/build. No review-work or Cubic."}
+        ]
+      }
+    },
+    {
+      "eval_name": "regex-fix-false-positive",
+      "with_skill": {
+        "pass_rate": 0.8,
+        "passed": 4,
+        "total": 5,
+        "duration_seconds": 570,
+        "failed_assertions": [
+          {"assertion": "Only modifies regex and adds tests — no unrelated changes", "reason": "Also proposes config schema change (exclude_patterns) and Go binary update — goes beyond minimal fix"}
+        ]
+      },
+      "without_skill": {
+        "pass_rate": 0.6,
+        "passed": 3,
+        "total": 5,
+        "duration_seconds": 399,
+        "failed_assertions": [
+          {"assertion": "Plan uses git worktree in a sibling directory", "reason": "git checkout -b, no worktree"},
+          {"assertion": "Verification loop includes all 3 gates", "reason": "Only bun test and typecheck. No review-work or Cubic."}
+        ]
+      }
+    }
+  ],
+  "analyst_observations": [
+    "Three-gates assertion (CI + review-work + Cubic) is the strongest discriminator: 5/5 with-skill vs 0/5 without-skill. Without the skill, agents never know about Cubic or review-work gates.",
+    "Worktree isolation is nearly as discriminating (5/5 vs 1/5). One without-skill run (eval-4) independently chose worktree, suggesting some agents already know worktree patterns, but the skill makes it consistent.",
+    "The skill's only failure (eval-5 minimal-change) reveals a potential over-engineering tendency: the skill-guided agent proposed config schema changes and Go binary updates for what should have been a minimal regex fix. Consider adding explicit guidance for fix-type tasks to stay minimal.",
+    "Duration tradeoff: with-skill is 12% slower on average (340s vs 303s), driven mainly by eval-2 (bugfix) and eval-5 (regex fix) where the skill's thorough verification planning adds overhead. For eval-1 and eval-3-4, with-skill was actually faster.",
+    "Without-skill duration has lower variance (stddev 78s vs 169s), suggesting the skill introduces more variable execution paths depending on task complexity.",
+    "Non-discriminating assertions: 'References actual files', 'PR targets dev', 'Runs local checks' — these pass regardless of skill. They validate baseline agent competence, not skill value. Consider removing or downweighting in future iterations.",
+    "Atomic commits assertion discriminates moderately (2/2 with-skill tested vs 0/2 without-skill tested). Without the skill, agents default to single commits even for multi-file refactors."
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/benchmark.md

@@ -0,0 +1,42 @@
+# Benchmark: work-with-pr (Iteration 1)
+
+## Summary
+
+| Metric | With Skill | Without Skill | Delta |
+|--------|-----------|---------------|-------|
+| Pass Rate | 96.8% (30/31) | 51.6% (16/31) | +45.2% |
+| Mean Duration | 340.2s | 303.0s | +37.2s |
+| Duration Stddev | 169.3s | 77.8s | +91.5s |
+
+## Per-Eval Breakdown
+
+| Eval | With Skill | Without Skill | Delta |
+|------|-----------|---------------|-------|
+| happy-path-feature-config-option | 100% (10/10) | 40% (4/10) | +60% |
+| bugfix-atlas-null-check | 100% (6/6) | 67% (4/6) | +33% |
+| refactor-split-constants | 100% (5/5) | 40% (2/5) | +60% |
+| new-mcp-arxiv-casual | 100% (5/5) | 60% (3/5) | +40% |
+| regex-fix-false-positive | 80% (4/5) | 60% (3/5) | +20% |
+
+## Key Discriminators
+
+- **three-gates** (CI + review-work + Cubic): 5/5 vs 0/5 — strongest signal
+- **worktree-isolation**: 5/5 vs 1/5
+- **atomic-commits**: 2/2 vs 0/2
+- **cubic-check-method**: 1/1 vs 0/1
+
+## Non-Discriminating Assertions
+
+- References actual files: passes in both conditions
+- PR targets dev: passes in both conditions
+- Runs local checks before pushing: passes in both conditions
+
+## Only With-Skill Failure
+
+- **eval-5 minimal-change**: Skill-guided agent proposed config schema changes and Go binary update for a minimal regex fix. The skill may encourage over-engineering in fix scenarios.
+
+## Analyst Notes
+
+- The skill adds most value for procedural knowledge (verification gates, worktree workflow) that agents cannot infer from codebase alone.
+- Duration cost is modest (+12%) and acceptable given the +45% pass rate improvement.
+- Consider adding explicit "fix-type tasks: stay minimal" guidance in iteration 2.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/eval_metadata.json

@@ -0,0 +1,57 @@
+{
+  "eval_id": 1,
+  "eval_name": "happy-path-feature-config-option",
+  "prompt": "I need to add a `max_background_agents` config option to oh-my-opencode that limits how many background agents can run simultaneously. It should be in the plugin config schema with a default of 5. Add validation and make sure the background manager respects it. Create a PR for this.",
+  "assertions": [
+    {
+      "id": "worktree-isolation",
+      "text": "Plan uses git worktree in a sibling directory (not main working directory)",
+      "type": "manual"
+    },
+    {
+      "id": "branch-from-dev",
+      "text": "Branch is created from origin/dev (not master/main)",
+      "type": "manual"
+    },
+    {
+      "id": "atomic-commits",
+      "text": "Plan specifies multiple atomic commits for multi-file changes",
+      "type": "manual"
+    },
+    {
+      "id": "local-validation",
+      "text": "Runs bun run typecheck, bun test, and bun run build before pushing",
+      "type": "manual"
+    },
+    {
+      "id": "pr-targets-dev",
+      "text": "PR is created targeting dev branch (not master)",
+      "type": "manual"
+    },
+    {
+      "id": "three-gates",
+      "text": "Verification loop includes all 3 gates: CI, review-work, and Cubic",
+      "type": "manual"
+    },
+    {
+      "id": "gate-ordering",
+      "text": "Gates are checked in order: CI first, then review-work, then Cubic",
+      "type": "manual"
+    },
+    {
+      "id": "cubic-check-method",
+      "text": "Cubic check uses gh api to check cubic-dev-ai[bot] reviews for 'No issues found'",
+      "type": "manual"
+    },
+    {
+      "id": "worktree-cleanup",
+      "text": "Plan includes worktree cleanup after merge",
+      "type": "manual"
+    },
+    {
+      "id": "real-file-references",
+      "text": "Code changes reference actual files in the codebase (config schema, background manager)",
+      "type": "manual"
+    }
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/with_skill/grading.json

@@ -0,0 +1,15 @@
+{
+  "run_id": "eval-1-with_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": true, "evidence": "Uses ../omo-wt/feat-max-background-agents"},
+    {"text": "Branch is created from origin/dev", "passed": true, "evidence": "git checkout dev && git pull origin dev, then branch"},
+    {"text": "Plan specifies multiple atomic commits for multi-file changes", "passed": true, "evidence": "2 commits: schema+tests, then concurrency+manager"},
+    {"text": "Runs bun run typecheck, bun test, and bun run build before pushing", "passed": true, "evidence": "Explicit pre-push section with all 3 commands"},
+    {"text": "PR is created targeting dev branch", "passed": true, "evidence": "--base dev in gh pr create"},
+    {"text": "Verification loop includes all 3 gates: CI, review-work, and Cubic", "passed": true, "evidence": "Gate A (CI), Gate B (review-work 5 agents), Gate C (Cubic)"},
+    {"text": "Gates are checked in order: CI first, then review-work, then Cubic", "passed": true, "evidence": "Explicit ordering in verify loop pseudocode"},
+    {"text": "Cubic check uses gh api to check cubic-dev-ai[bot] reviews", "passed": true, "evidence": "Mentions cubic-dev-ai[bot] and 'No issues found' signal"},
+    {"text": "Plan includes worktree cleanup after merge", "passed": true, "evidence": "Phase 4: git worktree remove ../omo-wt/feat-max-background-agents"},
+    {"text": "Code changes reference actual files in the codebase", "passed": true, "evidence": "References src/config/schema/background-task.ts, src/features/background-agent/concurrency.ts, manager.ts"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/with_skill/outputs/code-changes.md

@@ -0,0 +1,454 @@
+# Code Changes: `max_background_agents` Config Option
+
+## 1. `src/config/schema/background-task.ts` — Add schema field
+
+```typescript
+import { z } from "zod"
+
+export const BackgroundTaskConfigSchema = z.object({
+  defaultConcurrency: z.number().min(1).optional(),
+  providerConcurrency: z.record(z.string(), z.number().min(0)).optional(),
+  modelConcurrency: z.record(z.string(), z.number().min(0)).optional(),
+  maxDepth: z.number().int().min(1).optional(),
+  maxDescendants: z.number().int().min(1).optional(),
+  /** Maximum number of background agents that can run simultaneously across all models/providers (default: 5, minimum: 1) */
+  maxBackgroundAgents: z.number().int().min(1).optional(),
+  /** Stale timeout in milliseconds - interrupt tasks with no activity for this duration (default: 180000 = 3 minutes, minimum: 60000 = 1 minute) */
+  staleTimeoutMs: z.number().min(60000).optional(),
+  /** Timeout for tasks that never received any progress update, falling back to startedAt (default: 1800000 = 30 minutes, minimum: 60000 = 1 minute) */
+  messageStalenessTimeoutMs: z.number().min(60000).optional(),
+  syncPollTimeoutMs: z.number().min(60000).optional(),
+})
+
+export type BackgroundTaskConfig = z.infer<typeof BackgroundTaskConfigSchema>
+```
+
+**Rationale:** Follows exact same pattern as `maxDepth` and `maxDescendants` — `z.number().int().min(1).optional()`. The field is optional; runtime default of 5 is applied in `ConcurrencyManager`. No barrel export changes needed since `src/config/schema.ts` already does `export * from "./schema/background-task"` and the type is inferred.
+
+---
+
+## 2. `src/config/schema/background-task.test.ts` — Add validation tests
+
+Append after the existing `syncPollTimeoutMs` describe block (before the closing `})`):
+
+```typescript
+  describe("maxBackgroundAgents", () => {
+    describe("#given valid maxBackgroundAgents (10)", () => {
+      test("#when parsed #then returns correct value", () => {
+        const result = BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 10 })
+
+        expect(result.maxBackgroundAgents).toBe(10)
+      })
+    })
+
+    describe("#given maxBackgroundAgents of 1 (minimum)", () => {
+      test("#when parsed #then returns correct value", () => {
+        const result = BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 1 })
+
+        expect(result.maxBackgroundAgents).toBe(1)
+      })
+    })
+
+    describe("#given maxBackgroundAgents below minimum (0)", () => {
+      test("#when parsed #then throws ZodError", () => {
+        let thrownError: unknown
+
+        try {
+          BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 0 })
+        } catch (error) {
+          thrownError = error
+        }
+
+        expect(thrownError).toBeInstanceOf(ZodError)
+      })
+    })
+
+    describe("#given maxBackgroundAgents not provided", () => {
+      test("#when parsed #then field is undefined", () => {
+        const result = BackgroundTaskConfigSchema.parse({})
+
+        expect(result.maxBackgroundAgents).toBeUndefined()
+      })
+    })
+
+    describe('#given maxBackgroundAgents is non-integer (2.5)', () => {
+      test("#when parsed #then throws ZodError", () => {
+        let thrownError: unknown
+
+        try {
+          BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 2.5 })
+        } catch (error) {
+          thrownError = error
+        }
+
+        expect(thrownError).toBeInstanceOf(ZodError)
+      })
+    })
+  })
+```
+
+**Rationale:** Follows exact test pattern from `maxDepth`, `maxDescendants`, and `syncPollTimeoutMs` tests. Uses `#given`/`#when`/`#then` nested describe style. Tests valid, minimum boundary, below minimum, not provided, and non-integer cases.
+
+---
+
+## 3. `src/features/background-agent/concurrency.ts` — Add global agent limit
+
+```typescript
+import type { BackgroundTaskConfig } from "../../config/schema"
+
+const DEFAULT_MAX_BACKGROUND_AGENTS = 5
+
+/**
+ * Queue entry with settled-flag pattern to prevent double-resolution.
+ *
+ * The settled flag ensures that cancelWaiters() doesn't reject
+ * an entry that was already resolved by release().
+ */
+interface QueueEntry {
+  resolve: () => void
+  rawReject: (error: Error) => void
+  settled: boolean
+}
+
+export class ConcurrencyManager {
+  private config?: BackgroundTaskConfig
+  private counts: Map<string, number> = new Map()
+  private queues: Map<string, QueueEntry[]> = new Map()
+  private globalRunningCount = 0
+
+  constructor(config?: BackgroundTaskConfig) {
+    this.config = config
+  }
+
+  getMaxBackgroundAgents(): number {
+    return this.config?.maxBackgroundAgents ?? DEFAULT_MAX_BACKGROUND_AGENTS
+  }
+
+  getGlobalRunningCount(): number {
+    return this.globalRunningCount
+  }
+
+  canSpawnGlobally(): boolean {
+    return this.globalRunningCount < this.getMaxBackgroundAgents()
+  }
+
+  acquireGlobal(): void {
+    this.globalRunningCount++
+  }
+
+  releaseGlobal(): void {
+    if (this.globalRunningCount > 0) {
+      this.globalRunningCount--
+    }
+  }
+
+  getConcurrencyLimit(model: string): number {
+    // ... existing implementation unchanged ...
+  }
+
+  async acquire(model: string): Promise<void> {
+    // ... existing implementation unchanged ...
+  }
+
+  release(model: string): void {
+    // ... existing implementation unchanged ...
+  }
+
+  cancelWaiters(model: string): void {
+    // ... existing implementation unchanged ...
+  }
+
+  clear(): void {
+    for (const [model] of this.queues) {
+      this.cancelWaiters(model)
+    }
+    this.counts.clear()
+    this.queues.clear()
+    this.globalRunningCount = 0
+  }
+
+  getCount(model: string): number {
+    return this.counts.get(model) ?? 0
+  }
+
+  getQueueLength(model: string): number {
+    return this.queues.get(model)?.length ?? 0
+  }
+}
+```
+
+**Key changes:**
+- Add `DEFAULT_MAX_BACKGROUND_AGENTS = 5` constant
+- Add `globalRunningCount` private field
+- Add `getMaxBackgroundAgents()`, `getGlobalRunningCount()`, `canSpawnGlobally()`, `acquireGlobal()`, `releaseGlobal()` methods
+- `clear()` resets `globalRunningCount` to 0
+- All existing per-model methods remain unchanged
+
+---
+
+## 4. `src/features/background-agent/concurrency.test.ts` — Add global limit tests
+
+Append new describe block:
+
+```typescript
+describe("ConcurrencyManager global background agent limit", () => {
+  test("should default max background agents to 5 when no config", () => {
+    // given
+    const manager = new ConcurrencyManager()
+
+    // when
+    const max = manager.getMaxBackgroundAgents()
+
+    // then
+    expect(max).toBe(5)
+  })
+
+  test("should use configured maxBackgroundAgents", () => {
+    // given
+    const config: BackgroundTaskConfig = { maxBackgroundAgents: 10 }
+    const manager = new ConcurrencyManager(config)
+
+    // when
+    const max = manager.getMaxBackgroundAgents()
+
+    // then
+    expect(max).toBe(10)
+  })
+
+  test("should allow spawning when under global limit", () => {
+    // given
+    const config: BackgroundTaskConfig = { maxBackgroundAgents: 2 }
+    const manager = new ConcurrencyManager(config)
+
+    // when
+    manager.acquireGlobal()
+
+    // then
+    expect(manager.canSpawnGlobally()).toBe(true)
+    expect(manager.getGlobalRunningCount()).toBe(1)
+  })
+
+  test("should block spawning when at global limit", () => {
+    // given
+    const config: BackgroundTaskConfig = { maxBackgroundAgents: 2 }
+    const manager = new ConcurrencyManager(config)
+
+    // when
+    manager.acquireGlobal()
+    manager.acquireGlobal()
+
+    // then
+    expect(manager.canSpawnGlobally()).toBe(false)
+    expect(manager.getGlobalRunningCount()).toBe(2)
+  })
+
+  test("should allow spawning again after release", () => {
+    // given
+    const config: BackgroundTaskConfig = { maxBackgroundAgents: 1 }
+    const manager = new ConcurrencyManager(config)
+    manager.acquireGlobal()
+
+    // when
+    manager.releaseGlobal()
+
+    // then
+    expect(manager.canSpawnGlobally()).toBe(true)
+    expect(manager.getGlobalRunningCount()).toBe(0)
+  })
+
+  test("should not go below zero on extra release", () => {
+    // given
+    const manager = new ConcurrencyManager()
+
+    // when
+    manager.releaseGlobal()
+
+    // then
+    expect(manager.getGlobalRunningCount()).toBe(0)
+  })
+
+  test("should reset global count on clear", () => {
+    // given
+    const config: BackgroundTaskConfig = { maxBackgroundAgents: 5 }
+    const manager = new ConcurrencyManager(config)
+    manager.acquireGlobal()
+    manager.acquireGlobal()
+    manager.acquireGlobal()
+
+    // when
+    manager.clear()
+
+    // then
+    expect(manager.getGlobalRunningCount()).toBe(0)
+  })
+})
+```
+
+---
+
+## 5. `src/features/background-agent/manager.ts` — Enforce global limit
+
+### In `launch()` method — add check before task creation (after `reserveSubagentSpawn`):
+
+```typescript
+  async launch(input: LaunchInput): Promise<BackgroundTask> {
+    // ... existing logging ...
+
+    if (!input.agent || input.agent.trim() === "") {
+      throw new Error("Agent parameter is required")
+    }
+
+    // Check global background agent limit before spawn guard
+    if (!this.concurrencyManager.canSpawnGlobally()) {
+      const max = this.concurrencyManager.getMaxBackgroundAgents()
+      const current = this.concurrencyManager.getGlobalRunningCount()
+      throw new Error(
+        `Background agent spawn blocked: ${current} agents running, max is ${max}. Wait for existing tasks to complete or increase background_task.maxBackgroundAgents.`
+      )
+    }
+
+    const spawnReservation = await this.reserveSubagentSpawn(input.parentSessionID)
+
+    try {
+      // ... existing code ...
+
+      // After task creation, before queueing:
+      this.concurrencyManager.acquireGlobal()
+
+      // ... rest of existing code ...
+    } catch (error) {
+      spawnReservation.rollback()
+      throw error
+    }
+  }
+```
+
+### In `trackTask()` method — add global check:
+
+```typescript
+  async trackTask(input: { ... }): Promise<BackgroundTask> {
+    const existingTask = this.tasks.get(input.taskId)
+    if (existingTask) {
+      // ... existing re-registration logic unchanged ...
+      return existingTask
+    }
+
+    // Check global limit for new external tasks
+    if (!this.concurrencyManager.canSpawnGlobally()) {
+      const max = this.concurrencyManager.getMaxBackgroundAgents()
+      const current = this.concurrencyManager.getGlobalRunningCount()
+      throw new Error(
+        `Background agent spawn blocked: ${current} agents running, max is ${max}. Wait for existing tasks to complete or increase background_task.maxBackgroundAgents.`
+      )
+    }
+
+    // ... existing task creation ...
+    this.concurrencyManager.acquireGlobal()
+
+    // ... rest unchanged ...
+  }
+```
+
+### In `tryCompleteTask()` — release global slot:
+
+```typescript
+  private async tryCompleteTask(task: BackgroundTask, source: string): Promise<boolean> {
+    if (task.status !== "running") {
+      // ... existing guard ...
+      return false
+    }
+
+    task.status = "completed"
+    task.completedAt = new Date()
+    // ... existing history record ...
+
+    removeTaskToastTracking(task.id)
+
+    // Release per-model concurrency
+    if (task.concurrencyKey) {
+      this.concurrencyManager.release(task.concurrencyKey)
+      task.concurrencyKey = undefined
+    }
+
+    // Release global slot
+    this.concurrencyManager.releaseGlobal()
+
+    // ... rest unchanged ...
+  }
+```
+
+### In `cancelTask()` — release global slot:
+
+```typescript
+  async cancelTask(taskId: string, options?: { ... }): Promise<boolean> {
+    // ... existing code up to concurrency release ...
+
+    if (task.concurrencyKey) {
+      this.concurrencyManager.release(task.concurrencyKey)
+      task.concurrencyKey = undefined
+    }
+
+    // Release global slot (only for running tasks, pending never acquired)
+    if (task.status !== "pending") {
+      this.concurrencyManager.releaseGlobal()
+    }
+
+    // ... rest unchanged ...
+  }
+```
+
+### In `handleEvent()` session.error handler — release global slot:
+
+```typescript
+    if (event.type === "session.error") {
+      // ... existing error handling ...
+
+      task.status = "error"
+      // ...
+
+      if (task.concurrencyKey) {
+        this.concurrencyManager.release(task.concurrencyKey)
+        task.concurrencyKey = undefined
+      }
+
+      // Release global slot
+      this.concurrencyManager.releaseGlobal()
+
+      // ... rest unchanged ...
+    }
+```
+
+### In prompt error handler inside `startTask()` — release global slot:
+
+```typescript
+    promptWithModelSuggestionRetry(this.client, { ... }).catch((error) => {
+      // ... existing error handling ...
+      if (existingTask) {
+        existingTask.status = "interrupt"
+        // ...
+        if (existingTask.concurrencyKey) {
+          this.concurrencyManager.release(existingTask.concurrencyKey)
+          existingTask.concurrencyKey = undefined
+        }
+
+        // Release global slot
+        this.concurrencyManager.releaseGlobal()
+
+        // ... rest unchanged ...
+      }
+    })
+```
+
+---
+
+## Summary of Changes
+
+| File | Lines Added | Lines Modified |
+|------|-------------|----------------|
+| `src/config/schema/background-task.ts` | 2 | 0 |
+| `src/config/schema/background-task.test.ts` | ~50 | 0 |
+| `src/features/background-agent/concurrency.ts` | ~25 | 1 (`clear()`) |
+| `src/features/background-agent/concurrency.test.ts` | ~70 | 0 |
+| `src/features/background-agent/manager.ts` | ~20 | 0 |
+
+Total: ~167 lines added, 1 line modified across 5 files.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/with_skill/outputs/execution-plan.md

@@ -0,0 +1,136 @@
+# Execution Plan: `max_background_agents` Config Option
+
+## Phase 0: Setup — Branch + Worktree
+
+1. **Create branch** from `dev`:
+   ```bash
+   git checkout dev && git pull origin dev
+   git checkout -b feat/max-background-agents
+   ```
+
+2. **Create worktree** in sibling directory:
+   ```bash
+   mkdir -p ../omo-wt
+   git worktree add ../omo-wt/feat-max-background-agents feat/max-background-agents
+   ```
+
+3. **All subsequent work** happens in `../omo-wt/feat-max-background-agents/`, never in the main worktree.
+
+---
+
+## Phase 1: Implement — Atomic Commits
+
+### Commit 1: Add `max_background_agents` to config schema
+
+**Files changed:**
+- `src/config/schema/background-task.ts` — Add `maxBackgroundAgents` field to `BackgroundTaskConfigSchema`
+- `src/config/schema/background-task.test.ts` — Add validation tests for the new field
+
+**What:**
+- Add `maxBackgroundAgents: z.number().int().min(1).optional()` to `BackgroundTaskConfigSchema`
+- Default value handled at runtime (5), not in schema (all schema fields are optional per convention)
+- Add given/when/then tests: valid value, below minimum, not provided, non-number
+
+### Commit 2: Enforce limit in BackgroundManager + ConcurrencyManager
+
+**Files changed:**
+- `src/features/background-agent/concurrency.ts` — Add global agent count tracking + `getGlobalRunningCount()` + `canSpawnGlobally()`
+- `src/features/background-agent/concurrency.test.ts` — Tests for global limit enforcement
+- `src/features/background-agent/manager.ts` — Check global limit before `launch()` and `trackTask()`
+
+**What:**
+- `ConcurrencyManager` already manages per-model concurrency. Add a separate global counter:
+  - `private globalRunningCount: number = 0`
+  - `private maxBackgroundAgents: number` (from config, default 5)
+  - `acquireGlobal()` / `releaseGlobal()` methods
+  - `getGlobalRunningCount()` for observability
+- `BackgroundManager.launch()` checks `concurrencyManager.canSpawnGlobally()` before creating task
+- `BackgroundManager.trackTask()` also checks global limit
+- On task completion/cancellation/error, call `releaseGlobal()`
+- Throw descriptive error when limit hit: `"Background agent spawn blocked: ${current} agents running, max is ${max}. Wait for existing tasks to complete or increase background_task.maxBackgroundAgents."`
+
+### Local Validation
+
+```bash
+bun run typecheck
+bun test src/config/schema/background-task.test.ts
+bun test src/features/background-agent/concurrency.test.ts
+bun run build
+```
+
+---
+
+## Phase 2: PR Creation
+
+1. **Push branch:**
+   ```bash
+   git push -u origin feat/max-background-agents
+   ```
+
+2. **Create PR** targeting `dev`:
+   ```bash
+   gh pr create \
+     --base dev \
+     --title "feat: add max_background_agents config to limit concurrent background agents" \
+     --body-file /tmp/pull-request-max-background-agents-$(date +%s).md
+   ```
+
+---
+
+## Phase 3: Verify Loop
+
+### Gate A: CI
+- Wait for `ci.yml` workflow to complete
+- Check: `gh pr checks <PR_NUMBER> --watch`
+- If fails: read logs, fix, push, re-check
+
+### Gate B: review-work (5 agents)
+- Run `/review-work` skill which launches 5 parallel background sub-agents:
+  1. Oracle — goal/constraint verification
+  2. Oracle — code quality
+  3. Oracle — security
+  4. Hephaestus — hands-on QA execution
+  5. Hephaestus — context mining from GitHub/git
+- All 5 must pass. If any fails, fix and re-push.
+
+### Gate C: Cubic (cubic-dev-ai[bot])
+- Wait for Cubic bot review on PR
+- Must say "No issues found"
+- If issues found: address feedback, push, re-check
+
+### Loop
+```
+while (!allGatesPass) {
+  if (CI fails) → fix → push → continue
+  if (review-work fails) → fix → push → continue
+  if (Cubic has issues) → fix → push → continue
+}
+```
+
+---
+
+## Phase 4: Merge + Cleanup
+
+1. **Squash merge:**
+   ```bash
+   gh pr merge <PR_NUMBER> --squash --delete-branch
+   ```
+
+2. **Remove worktree:**
+   ```bash
+   git worktree remove ../omo-wt/feat-max-background-agents
+   ```
+
+---
+
+## File Impact Summary
+
+| File | Change Type |
+|------|-------------|
+| `src/config/schema/background-task.ts` | Modified — add schema field |
+| `src/config/schema/background-task.test.ts` | Modified — add validation tests |
+| `src/features/background-agent/concurrency.ts` | Modified — add global limit tracking |
+| `src/features/background-agent/concurrency.test.ts` | Modified — add global limit tests |
+| `src/features/background-agent/manager.ts` | Modified — enforce global limit in launch/trackTask |
+
+5 files changed across 2 atomic commits. No new files created (follows existing patterns).

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/with_skill/outputs/pr-description.md

@@ -0,0 +1,47 @@
+# PR Description
+
+**Title:** `feat: add max_background_agents config to limit concurrent background agents`
+
+**Base:** `dev`
+
+---
+
+## Summary
+
+- Add `maxBackgroundAgents` field to `BackgroundTaskConfigSchema` (default: 5, min: 1) to cap total simultaneous background agents across all models/providers
+- Enforce the global limit in `BackgroundManager.launch()` and `trackTask()` with descriptive error messages when the limit is hit
+- Release global slots on task completion, cancellation, error, and interrupt to prevent slot leaks
+
+## Motivation
+
+The existing concurrency system in `ConcurrencyManager` limits agents **per model/provider** (e.g., 5 concurrent `anthropic/claude-opus-4-6` tasks). However, there is no **global** cap across all models. A user running tasks across multiple providers could spawn an unbounded number of background agents, exhausting system resources.
+
+`max_background_agents` provides a single knob to limit total concurrent background agents regardless of which model they use.
+
+## Config Usage
+
+```jsonc
+// .opencode/oh-my-opencode.jsonc
+{
+  "background_task": {
+    "maxBackgroundAgents": 10  // default: 5, min: 1
+  }
+}
+```
+
+## Changes
+
+| File | What |
+|------|------|
+| `src/config/schema/background-task.ts` | Add `maxBackgroundAgents` schema field |
+| `src/config/schema/background-task.test.ts` | Validation tests (valid, boundary, invalid) |
+| `src/features/background-agent/concurrency.ts` | Global counter + `canSpawnGlobally()` / `acquireGlobal()` / `releaseGlobal()` |
+| `src/features/background-agent/concurrency.test.ts` | Global limit unit tests |
+| `src/features/background-agent/manager.ts` | Enforce global limit in `launch()`, `trackTask()`; release in completion/cancel/error paths |
+
+## Testing
+
+- `bun test src/config/schema/background-task.test.ts` — schema validation
+- `bun test src/features/background-agent/concurrency.test.ts` — global limit enforcement
+- `bun run typecheck` — clean
+- `bun run build` — clean

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/with_skill/outputs/verification-strategy.md

@@ -0,0 +1,163 @@
+# Verification Strategy
+
+## Pre-Push Local Validation
+
+Before every push, run all three checks sequentially:
+
+```bash
+bun run typecheck && bun test && bun run build
+```
+
+Specific test files to watch:
+```bash
+bun test src/config/schema/background-task.test.ts
+bun test src/features/background-agent/concurrency.test.ts
+```
+
+---
+
+## Gate A: CI (`ci.yml`)
+
+### What CI runs
+1. **Tests (split):** mock-heavy tests run in isolation (separate `bun test` processes), rest in batch
+2. **Typecheck:** `bun run typecheck` (tsc --noEmit)
+3. **Build:** `bun run build` (ESM + declarations + schema)
+4. **Schema auto-commit:** if generated schema changed, CI commits it
+
+### How to monitor
+```bash
+gh pr checks <PR_NUMBER> --watch
+```
+
+### Common failure scenarios and fixes
+
+| Failure | Likely Cause | Fix |
+|---------|-------------|-----|
+| Typecheck error | New field not matching existing type imports | Verify `BackgroundTaskConfig` type is auto-inferred from schema, no manual type updates needed |
+| Test failure | Test assertion wrong or missing import | Fix test, re-push |
+| Build failure | Import cycle or missing export | Check barrel exports in `src/config/schema.ts` (already re-exports via `export *`) |
+| Schema auto-commit | Generated JSON schema changed | Pull the auto-commit, rebase if needed |
+
+### Recovery
+```bash
+# Read CI logs
+gh run view <RUN_ID> --log-failed
+
+# Fix, commit, push
+git add -A && git commit -m "fix: address CI failure" && git push
+```
+
+---
+
+## Gate B: review-work (5 parallel agents)
+
+### What it checks
+Run `/review-work` which launches 5 background sub-agents:
+
+| Agent | Role | What it checks for this PR |
+|-------|------|---------------------------|
+| Oracle (goal) | Goal/constraint verification | Does `maxBackgroundAgents` actually limit agents? Is default 5? Is min 1? |
+| Oracle (quality) | Code quality | Follows existing patterns? No catch-all files? Under 200 LOC? given/when/then tests? |
+| Oracle (security) | Security review | No injection vectors, no unsafe defaults, proper input validation via Zod |
+| Hephaestus (QA) | Hands-on QA execution | Actually runs tests, checks typecheck, verifies build |
+| Hephaestus (context) | Context mining | Checks git history, related issues, ensures no duplicate/conflicting PRs |
+
+### Pass criteria
+All 5 agents must pass. Any single failure blocks.
+
+### Common failure scenarios and fixes
+
+| Agent | Likely Issue | Fix |
+|-------|-------------|-----|
+| Oracle (goal) | Global limit not enforced in all exit paths (completion, cancel, error, interrupt) | Audit every status transition in `manager.ts` that should call `releaseGlobal()` |
+| Oracle (quality) | Test style not matching given/when/then | Restructure tests with `#given`/`#when`/`#then` describe nesting |
+| Oracle (quality) | File exceeds 200 LOC | `concurrency.ts` is 137 LOC + ~25 new = ~162 LOC, safe. `manager.ts` is already large but we're adding ~20 lines to existing methods, not creating new responsibility |
+| Oracle (security) | Integer overflow or negative values | Zod `.int().min(1)` handles this at config parse time |
+| Hephaestus (QA) | Test actually fails when run | Run tests locally first, fix before push |
+
+### Recovery
+```bash
+# Review agent output
+background_output(task_id="<review-work-task-id>")
+
+# Fix identified issues
+# ... edit files ...
+git add -A && git commit -m "fix: address review-work feedback" && git push
+```
+
+---
+
+## Gate C: Cubic (`cubic-dev-ai[bot]`)
+
+### What it checks
+Cubic is an automated code review bot that analyzes the PR diff. It must respond with "No issues found" for the gate to pass.
+
+### Common failure scenarios and fixes
+
+| Issue | Likely Cause | Fix |
+|-------|-------------|-----|
+| "Missing error handling" | `releaseGlobal()` not called in some error path | Add `releaseGlobal()` to the missed path |
+| "Inconsistent naming" | Field name doesn't match convention | Use `maxBackgroundAgents` (camelCase in schema, `max_background_agents` in JSONC config) |
+| "Missing documentation" | No JSDoc on new public methods | Add JSDoc comments to `canSpawnGlobally()`, `acquireGlobal()`, `releaseGlobal()`, `getMaxBackgroundAgents()` |
+| "Test coverage gap" | Missing edge case test | Add the specific test case Cubic identifies |
+
+### Recovery
+```bash
+# Read Cubic's review
+gh api repos/code-yeongyu/oh-my-openagent/pulls/<PR_NUMBER>/reviews
+
+# Address each comment
+# ... edit files ...
+git add -A && git commit -m "fix: address Cubic review feedback" && git push
+```
+
+---
+
+## Verification Loop Pseudocode
+
+```
+iteration = 0
+while true:
+  iteration++
+  log("Verification iteration ${iteration}")
+
+  # Gate A: CI (cheapest, check first)
+  push_and_wait_for_ci()
+  if ci_failed:
+    read_ci_logs()
+    fix_and_commit()
+    continue
+
+  # Gate B: review-work (5 agents, more expensive)
+  run_review_work()
+  if any_agent_failed:
+    read_agent_feedback()
+    fix_and_commit()
+    continue
+
+  # Gate C: Cubic (external bot, wait for it)
+  wait_for_cubic_review()
+  if cubic_has_issues:
+    read_cubic_comments()
+    fix_and_commit()
+    continue
+
+  # All gates passed
+  break
+
+# Merge
+gh pr merge <PR_NUMBER> --squash --delete-branch
+```
+
+No iteration cap. Loop continues until all three gates pass simultaneously in a single iteration.
+
+---
+
+## Risk Assessment
+
+| Risk | Probability | Mitigation |
+|------|------------|------------|
+| Slot leak (global count never decremented) | Medium | Audit every exit path: `tryCompleteTask`, `cancelTask`, `handleEvent(session.error)`, `startTask` prompt error, `resume` prompt error |
+| Race condition on global count | Low | `globalRunningCount` is synchronous (single-threaded JS), no async gap between check and increment in `launch()` |
+| Breaking existing behavior | Low | Default is 5, same as existing per-model default. Users with <5 total agents see no change |
+| `manager.ts` exceeding 200 LOC | Already exceeded | File is already ~1500 LOC (exempt due to being a core orchestration class with many methods). Our changes add ~20 lines to existing methods, not a new responsibility |

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/with_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 292000, "total_duration_seconds": 292}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/without_skill/grading.json

@@ -0,0 +1,15 @@
+{
+  "run_id": "eval-1-without_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": false, "evidence": "Uses git checkout -b, no worktree isolation"},
+    {"text": "Branch is created from origin/dev", "passed": true, "evidence": "git checkout -b feat/max-background-agents dev"},
+    {"text": "Plan specifies multiple atomic commits for multi-file changes", "passed": false, "evidence": "Steps listed sequentially but no atomic commit strategy mentioned"},
+    {"text": "Runs bun run typecheck, bun test, and bun run build before pushing", "passed": true, "evidence": "Step 6 runs typecheck and tests, Step 8 implies push after verification"},
+    {"text": "PR is created targeting dev branch", "passed": true, "evidence": "Step 8 mentions creating PR"},
+    {"text": "Verification loop includes all 3 gates: CI, review-work, and Cubic", "passed": false, "evidence": "Only mentions CI pipeline in step 6. No review-work or Cubic."},
+    {"text": "Gates are checked in order: CI first, then review-work, then Cubic", "passed": false, "evidence": "No gate ordering - only CI mentioned"},
+    {"text": "Cubic check uses gh api to check cubic-dev-ai[bot] reviews", "passed": false, "evidence": "No mention of Cubic at all"},
+    {"text": "Plan includes worktree cleanup after merge", "passed": false, "evidence": "No worktree used, no cleanup needed"},
+    {"text": "Code changes reference actual files in the codebase", "passed": true, "evidence": "References actual files with detailed design decisions"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/without_skill/outputs/code-changes.md

@@ -0,0 +1,615 @@
+# Code Changes: `max_background_agents` Config Option
+
+## 1. Schema Change
+
+**File:** `src/config/schema/background-task.ts`
+
+```typescript
+import { z } from "zod"
+
+export const BackgroundTaskConfigSchema = z.object({
+  defaultConcurrency: z.number().min(1).optional(),
+  providerConcurrency: z.record(z.string(), z.number().min(0)).optional(),
+  modelConcurrency: z.record(z.string(), z.number().min(0)).optional(),
+  maxDepth: z.number().int().min(1).optional(),
+  maxDescendants: z.number().int().min(1).optional(),
+  /** Maximum number of background agents that can run simultaneously across all models/providers (default: no global limit, only per-model limits apply) */
+  maxBackgroundAgents: z.number().int().min(1).optional(),
+  /** Stale timeout in milliseconds - interrupt tasks with no activity for this duration (default: 180000 = 3 minutes, minimum: 60000 = 1 minute) */
+  staleTimeoutMs: z.number().min(60000).optional(),
+  /** Timeout for tasks that never received any progress update, falling back to startedAt (default: 1800000 = 30 minutes, minimum: 60000 = 1 minute) */
+  messageStalenessTimeoutMs: z.number().min(60000).optional(),
+  syncPollTimeoutMs: z.number().min(60000).optional(),
+})
+
+export type BackgroundTaskConfig = z.infer<typeof BackgroundTaskConfigSchema>
+```
+
+**What changed:** Added `maxBackgroundAgents` field after `maxDescendants` (grouped with other limit fields). Uses `z.number().int().min(1).optional()` matching the pattern of `maxDepth` and `maxDescendants`.
+
+---
+
+## 2. ConcurrencyManager Changes
+
+**File:** `src/features/background-agent/concurrency.ts`
+
+```typescript
+import type { BackgroundTaskConfig } from "../../config/schema"
+
+/**
+ * Queue entry with settled-flag pattern to prevent double-resolution.
+ *
+ * The settled flag ensures that cancelWaiters() doesn't reject
+ * an entry that was already resolved by release().
+ */
+interface QueueEntry {
+  resolve: () => void
+  rawReject: (error: Error) => void
+  settled: boolean
+}
+
+export class ConcurrencyManager {
+  private config?: BackgroundTaskConfig
+  private counts: Map<string, number> = new Map()
+  private queues: Map<string, QueueEntry[]> = new Map()
+  private globalCount = 0
+  private globalQueue: QueueEntry[] = []
+
+  constructor(config?: BackgroundTaskConfig) {
+    this.config = config
+  }
+
+  getGlobalLimit(): number {
+    const limit = this.config?.maxBackgroundAgents
+    if (limit === undefined) {
+      return Infinity
+    }
+    return limit
+  }
+
+  getConcurrencyLimit(model: string): number {
+    const modelLimit = this.config?.modelConcurrency?.[model]
+    if (modelLimit !== undefined) {
+      return modelLimit === 0 ? Infinity : modelLimit
+    }
+    const provider = model.split('/')[0]
+    const providerLimit = this.config?.providerConcurrency?.[provider]
+    if (providerLimit !== undefined) {
+      return providerLimit === 0 ? Infinity : providerLimit
+    }
+    const defaultLimit = this.config?.defaultConcurrency
+    if (defaultLimit !== undefined) {
+      return defaultLimit === 0 ? Infinity : defaultLimit
+    }
+    return 5
+  }
+
+  async acquire(model: string): Promise<void> {
+    const perModelLimit = this.getConcurrencyLimit(model)
+    const globalLimit = this.getGlobalLimit()
+
+    // Fast path: both limits have capacity
+    if (perModelLimit === Infinity && globalLimit === Infinity) {
+      return
+    }
+
+    const currentPerModel = this.counts.get(model) ?? 0
+
+    if (currentPerModel < perModelLimit && this.globalCount < globalLimit) {
+      this.counts.set(model, currentPerModel + 1)
+      this.globalCount++
+      return
+    }
+
+    return new Promise<void>((resolve, reject) => {
+      const entry: QueueEntry = {
+        resolve: () => {
+          if (entry.settled) return
+          entry.settled = true
+          resolve()
+        },
+        rawReject: reject,
+        settled: false,
+      }
+
+      // Queue on whichever limit is blocking
+      if (currentPerModel >= perModelLimit) {
+        const queue = this.queues.get(model) ?? []
+        queue.push(entry)
+        this.queues.set(model, queue)
+      } else {
+        this.globalQueue.push(entry)
+      }
+    })
+  }
+
+  release(model: string): void {
+    const perModelLimit = this.getConcurrencyLimit(model)
+    const globalLimit = this.getGlobalLimit()
+
+    if (perModelLimit === Infinity && globalLimit === Infinity) {
+      return
+    }
+
+    // Try per-model handoff first
+    const queue = this.queues.get(model)
+    while (queue && queue.length > 0) {
+      const next = queue.shift()!
+      if (!next.settled) {
+        // Hand off the slot to this waiter (counts stay the same)
+        next.resolve()
+        return
+      }
+    }
+
+    // No per-model handoff - decrement per-model count
+    const current = this.counts.get(model) ?? 0
+    if (current > 0) {
+      this.counts.set(model, current - 1)
+    }
+
+    // Try global handoff
+    while (this.globalQueue.length > 0) {
+      const next = this.globalQueue.shift()!
+      if (!next.settled) {
+        // Hand off the global slot - but the waiter still needs a per-model slot
+        // Since they were queued on global, their per-model had capacity
+        // Re-acquire per-model count for them
+        const waiterModel = this.findModelForGlobalWaiter()
+        if (waiterModel) {
+          const waiterCount = this.counts.get(waiterModel) ?? 0
+          this.counts.set(waiterModel, waiterCount + 1)
+        }
+        next.resolve()
+        return
+      }
+    }
+
+    // No handoff occurred - decrement global count
+    if (this.globalCount > 0) {
+      this.globalCount--
+    }
+  }
+
+  /**
+   * Cancel all waiting acquires for a model. Used during cleanup.
+   */
+  cancelWaiters(model: string): void {
+    const queue = this.queues.get(model)
+    if (queue) {
+      for (const entry of queue) {
+        if (!entry.settled) {
+          entry.settled = true
+          entry.rawReject(new Error(`Concurrency queue cancelled for model: ${model}`))
+        }
+      }
+      this.queues.delete(model)
+    }
+  }
+
+  /**
+   * Clear all state. Used during manager cleanup/shutdown.
+   * Cancels all pending waiters.
+   */
+  clear(): void {
+    for (const [model] of this.queues) {
+      this.cancelWaiters(model)
+    }
+    // Cancel global queue waiters
+    for (const entry of this.globalQueue) {
+      if (!entry.settled) {
+        entry.settled = true
+        entry.rawReject(new Error("Concurrency queue cancelled: manager shutdown"))
+      }
+    }
+    this.globalQueue = []
+    this.globalCount = 0
+    this.counts.clear()
+    this.queues.clear()
+  }
+
+  /**
+   * Get current count for a model (for testing/debugging)
+   */
+  getCount(model: string): number {
+    return this.counts.get(model) ?? 0
+  }
+
+  /**
+   * Get queue length for a model (for testing/debugging)
+   */
+  getQueueLength(model: string): number {
+    return this.queues.get(model)?.length ?? 0
+  }
+
+  /**
+   * Get current global count across all models (for testing/debugging)
+   */
+  getGlobalCount(): number {
+    return this.globalCount
+  }
+
+  /**
+   * Get global queue length (for testing/debugging)
+   */
+  getGlobalQueueLength(): number {
+    return this.globalQueue.length
+  }
+}
+```
+
+**What changed:**
+- Added `globalCount` field to track total active agents across all keys
+- Added `globalQueue` for tasks waiting on the global limit
+- Added `getGlobalLimit()` method to read `maxBackgroundAgents` from config
+- Modified `acquire()` to check both per-model AND global limits
+- Modified `release()` to handle global queue handoff and decrement global count
+- Modified `clear()` to reset global state
+- Added `getGlobalCount()` and `getGlobalQueueLength()` for testing
+
+**Important design note:** The `release()` implementation above is a simplified version. In practice, the global queue handoff is tricky because we need to know which model the global waiter was trying to acquire for. A cleaner approach would be to store the model key in the QueueEntry. Let me refine:
+
+### Refined approach (simpler, more correct)
+
+Instead of a separate global queue, a simpler approach is to check the global limit inside `acquire()` and use a single queue per model. When global capacity frees up on `release()`, we try to drain any model's queue:
+
+```typescript
+async acquire(model: string): Promise<void> {
+  const perModelLimit = this.getConcurrencyLimit(model)
+  const globalLimit = this.getGlobalLimit()
+
+  if (perModelLimit === Infinity && globalLimit === Infinity) {
+    return
+  }
+
+  const currentPerModel = this.counts.get(model) ?? 0
+
+  if (currentPerModel < perModelLimit && this.globalCount < globalLimit) {
+    this.counts.set(model, currentPerModel + 1)
+    if (globalLimit !== Infinity) {
+      this.globalCount++
+    }
+    return
+  }
+
+  return new Promise<void>((resolve, reject) => {
+    const queue = this.queues.get(model) ?? []
+
+    const entry: QueueEntry = {
+      resolve: () => {
+        if (entry.settled) return
+        entry.settled = true
+        resolve()
+      },
+      rawReject: reject,
+      settled: false,
+    }
+
+    queue.push(entry)
+    this.queues.set(model, queue)
+  })
+}
+
+release(model: string): void {
+  const perModelLimit = this.getConcurrencyLimit(model)
+  const globalLimit = this.getGlobalLimit()
+
+  if (perModelLimit === Infinity && globalLimit === Infinity) {
+    return
+  }
+
+  // Try per-model handoff first (same model queue)
+  const queue = this.queues.get(model)
+  while (queue && queue.length > 0) {
+    const next = queue.shift()!
+    if (!next.settled) {
+      // Hand off the slot to this waiter (per-model and global counts stay the same)
+      next.resolve()
+      return
+    }
+  }
+
+  // No per-model handoff - decrement per-model count
+  const current = this.counts.get(model) ?? 0
+  if (current > 0) {
+    this.counts.set(model, current - 1)
+  }
+
+  // Decrement global count
+  if (globalLimit !== Infinity && this.globalCount > 0) {
+    this.globalCount--
+  }
+
+  // Try to drain any other model's queue that was blocked by global limit
+  if (globalLimit !== Infinity) {
+    this.tryDrainGlobalWaiters()
+  }
+}
+
+private tryDrainGlobalWaiters(): void {
+  const globalLimit = this.getGlobalLimit()
+  if (this.globalCount >= globalLimit) return
+
+  for (const [model, queue] of this.queues) {
+    const perModelLimit = this.getConcurrencyLimit(model)
+    const currentPerModel = this.counts.get(model) ?? 0
+
+    if (currentPerModel >= perModelLimit) continue
+
+    while (queue.length > 0 && this.globalCount < globalLimit && currentPerModel < perModelLimit) {
+      const next = queue.shift()!
+      if (!next.settled) {
+        this.counts.set(model, (this.counts.get(model) ?? 0) + 1)
+        this.globalCount++
+        next.resolve()
+        return
+      }
+    }
+  }
+}
+```
+
+This refined approach keeps all waiters in per-model queues (no separate global queue), and on release, tries to drain waiters from any model queue that was blocked by the global limit.
+
+---
+
+## 3. Schema Test Changes
+
+**File:** `src/config/schema/background-task.test.ts`
+
+Add after the `syncPollTimeoutMs` describe block:
+
+```typescript
+  describe("maxBackgroundAgents", () => {
+    describe("#given valid maxBackgroundAgents (10)", () => {
+      test("#when parsed #then returns correct value", () => {
+        const result = BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 10 })
+
+        expect(result.maxBackgroundAgents).toBe(10)
+      })
+    })
+
+    describe("#given maxBackgroundAgents of 1 (minimum)", () => {
+      test("#when parsed #then returns correct value", () => {
+        const result = BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 1 })
+
+        expect(result.maxBackgroundAgents).toBe(1)
+      })
+    })
+
+    describe("#given maxBackgroundAgents below minimum (0)", () => {
+      test("#when parsed #then throws ZodError", () => {
+        let thrownError: unknown
+
+        try {
+          BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 0 })
+        } catch (error) {
+          thrownError = error
+        }
+
+        expect(thrownError).toBeInstanceOf(ZodError)
+      })
+    })
+
+    describe("#given maxBackgroundAgents is negative (-1)", () => {
+      test("#when parsed #then throws ZodError", () => {
+        let thrownError: unknown
+
+        try {
+          BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: -1 })
+        } catch (error) {
+          thrownError = error
+        }
+
+        expect(thrownError).toBeInstanceOf(ZodError)
+      })
+    })
+
+    describe("#given maxBackgroundAgents is non-integer (2.5)", () => {
+      test("#when parsed #then throws ZodError", () => {
+        let thrownError: unknown
+
+        try {
+          BackgroundTaskConfigSchema.parse({ maxBackgroundAgents: 2.5 })
+        } catch (error) {
+          thrownError = error
+        }
+
+        expect(thrownError).toBeInstanceOf(ZodError)
+      })
+    })
+
+    describe("#given maxBackgroundAgents not provided", () => {
+      test("#when parsed #then field is undefined", () => {
+        const result = BackgroundTaskConfigSchema.parse({})
+
+        expect(result.maxBackgroundAgents).toBeUndefined()
+      })
+    })
+  })
+```
+
+---
+
+## 4. ConcurrencyManager Test Changes
+
+**File:** `src/features/background-agent/concurrency.test.ts`
+
+Add new describe block:
+
+```typescript
+describe("ConcurrencyManager.globalLimit (maxBackgroundAgents)", () => {
+  test("should return Infinity when maxBackgroundAgents is not set", () => {
+    // given
+    const manager = new ConcurrencyManager()
+
+    // when
+    const limit = manager.getGlobalLimit()
+
+    // then
+    expect(limit).toBe(Infinity)
+  })
+
+  test("should return configured maxBackgroundAgents", () => {
+    // given
+    const config: BackgroundTaskConfig = { maxBackgroundAgents: 3 }
+    const manager = new ConcurrencyManager(config)
+
+    // when
+    const limit = manager.getGlobalLimit()
+
+    // then
+    expect(limit).toBe(3)
+  })
+
+  test("should enforce global limit across different models", async () => {
+    // given
+    const config: BackgroundTaskConfig = {
+      maxBackgroundAgents: 2,
+      defaultConcurrency: 5,
+    }
+    const manager = new ConcurrencyManager(config)
+    await manager.acquire("model-a")
+    await manager.acquire("model-b")
+
+    // when
+    let resolved = false
+    const waitPromise = manager.acquire("model-c").then(() => { resolved = true })
+    await Promise.resolve()
+
+    // then - should be blocked by global limit even though per-model has capacity
+    expect(resolved).toBe(false)
+    expect(manager.getGlobalCount()).toBe(2)
+
+    // cleanup
+    manager.release("model-a")
+    await waitPromise
+    expect(resolved).toBe(true)
+  })
+
+  test("should allow tasks when global limit not reached", async () => {
+    // given
+    const config: BackgroundTaskConfig = {
+      maxBackgroundAgents: 3,
+      defaultConcurrency: 5,
+    }
+    const manager = new ConcurrencyManager(config)
+
+    // when
+    await manager.acquire("model-a")
+    await manager.acquire("model-b")
+    await manager.acquire("model-c")
+
+    // then
+    expect(manager.getGlobalCount()).toBe(3)
+    expect(manager.getCount("model-a")).toBe(1)
+    expect(manager.getCount("model-b")).toBe(1)
+    expect(manager.getCount("model-c")).toBe(1)
+  })
+
+  test("should respect both per-model and global limits", async () => {
+    // given - per-model limit of 1, global limit of 3
+    const config: BackgroundTaskConfig = {
+      maxBackgroundAgents: 3,
+      defaultConcurrency: 1,
+    }
+    const manager = new ConcurrencyManager(config)
+    await manager.acquire("model-a")
+
+    // when - try second acquire on same model
+    let resolved = false
+    const waitPromise = manager.acquire("model-a").then(() => { resolved = true })
+    await Promise.resolve()
+
+    // then - blocked by per-model limit, not global
+    expect(resolved).toBe(false)
+    expect(manager.getGlobalCount()).toBe(1)
+
+    // cleanup
+    manager.release("model-a")
+    await waitPromise
+  })
+
+  test("should release global slot and unblock waiting tasks", async () => {
+    // given
+    const config: BackgroundTaskConfig = {
+      maxBackgroundAgents: 1,
+      defaultConcurrency: 5,
+    }
+    const manager = new ConcurrencyManager(config)
+    await manager.acquire("model-a")
+
+    // when
+    let resolved = false
+    const waitPromise = manager.acquire("model-b").then(() => { resolved = true })
+    await Promise.resolve()
+    expect(resolved).toBe(false)
+
+    manager.release("model-a")
+    await waitPromise
+
+    // then
+    expect(resolved).toBe(true)
+    expect(manager.getGlobalCount()).toBe(1)
+    expect(manager.getCount("model-a")).toBe(0)
+    expect(manager.getCount("model-b")).toBe(1)
+  })
+
+  test("should not enforce global limit when not configured", async () => {
+    // given - no maxBackgroundAgents set
+    const config: BackgroundTaskConfig = { defaultConcurrency: 5 }
+    const manager = new ConcurrencyManager(config)
+
+    // when - acquire many across different models
+    await manager.acquire("model-a")
+    await manager.acquire("model-b")
+    await manager.acquire("model-c")
+    await manager.acquire("model-d")
+    await manager.acquire("model-e")
+    await manager.acquire("model-f")
+
+    // then - all should succeed (no global limit)
+    expect(manager.getCount("model-a")).toBe(1)
+    expect(manager.getCount("model-f")).toBe(1)
+  })
+
+  test("should reset global count on clear", async () => {
+    // given
+    const config: BackgroundTaskConfig = { maxBackgroundAgents: 5 }
+    const manager = new ConcurrencyManager(config)
+    await manager.acquire("model-a")
+    await manager.acquire("model-b")
+
+    // when
+    manager.clear()
+
+    // then
+    expect(manager.getGlobalCount()).toBe(0)
+  })
+})
+```
+
+---
+
+## Config Usage Example
+
+User's `.opencode/oh-my-opencode.jsonc`:
+
+```jsonc
+{
+  "background_task": {
+    // Global limit: max 5 background agents total
+    "maxBackgroundAgents": 5,
+    // Per-model limits still apply independently
+    "defaultConcurrency": 3,
+    "providerConcurrency": {
+      "anthropic": 2
+    }
+  }
+}
+```
+
+With this config:
+- Max 5 background agents running simultaneously across all models
+- Max 3 per model (default), max 2 for any Anthropic model
+- If 2 Anthropic + 3 OpenAI agents are running (5 total), no more can start regardless of per-model capacity

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/without_skill/outputs/execution-plan.md

@@ -0,0 +1,99 @@
+# Execution Plan: Add `max_background_agents` Config Option
+
+## Overview
+
+Add a `max_background_agents` config option to oh-my-opencode that limits total simultaneous background agents across all models/providers. Currently, concurrency is only limited per-model/provider key (default 5 per key). This new option adds a **global ceiling** on total running background agents.
+
+## Step-by-Step Plan
+
+### Step 1: Create feature branch
+
+```bash
+git checkout -b feat/max-background-agents dev
+```
+
+### Step 2: Add `max_background_agents` to BackgroundTaskConfigSchema
+
+**File:** `src/config/schema/background-task.ts`
+
+- Add `maxBackgroundAgents` field to the Zod schema with `z.number().int().min(1).optional()`
+- This follows the existing pattern of `maxDepth` and `maxDescendants` (integer, min 1, optional)
+- The field name uses camelCase to match existing schema fields (`defaultConcurrency`, `maxDepth`, `maxDescendants`)
+- No `.default()` needed since the hardcoded fallback of 5 lives in `ConcurrencyManager`
+
+### Step 3: Modify `ConcurrencyManager` to enforce global limit
+
+**File:** `src/features/background-agent/concurrency.ts`
+
+- Add a `globalCount` field tracking total active agents across all keys
+- Modify `acquire()` to check global count against `maxBackgroundAgents` before granting a slot
+- Modify `release()` to decrement global count
+- Modify `clear()` to reset global count
+- Add `getGlobalCount()` for testing/debugging (follows existing `getCount()`/`getQueueLength()` pattern)
+
+The global limit check happens **in addition to** the per-model limit. Both must have capacity for a task to proceed.
+
+### Step 4: Add tests for the new config schema field
+
+**File:** `src/config/schema/background-task.test.ts`
+
+- Add test cases following the existing given/when/then pattern with nested describes
+- Test valid value, below-minimum value, undefined (not provided), non-number type
+
+### Step 5: Add tests for ConcurrencyManager global limit
+
+**File:** `src/features/background-agent/concurrency.test.ts`
+
+- Test that global limit is enforced across different model keys
+- Test that tasks queue when global limit reached even if per-model limit has capacity
+- Test that releasing a slot from one model allows a queued task from another model to proceed
+- Test default behavior (5) when no config provided
+- Test interaction between global and per-model limits
+
+### Step 6: Run typecheck and tests
+
+```bash
+bun run typecheck
+bun test src/config/schema/background-task.test.ts
+bun test src/features/background-agent/concurrency.test.ts
+```
+
+### Step 7: Verify LSP diagnostics clean
+
+Check `src/config/schema/background-task.ts` and `src/features/background-agent/concurrency.ts` for errors.
+
+### Step 8: Create PR
+
+- Push branch to remote
+- Create PR with structured description via `gh pr create`
+
+## Files Modified (4 files)
+
+| File | Change |
+|------|--------|
+| `src/config/schema/background-task.ts` | Add `maxBackgroundAgents` field |
+| `src/features/background-agent/concurrency.ts` | Add global count tracking + enforcement |
+| `src/config/schema/background-task.test.ts` | Add schema validation tests |
+| `src/features/background-agent/concurrency.test.ts` | Add global limit enforcement tests |
+
+## Files NOT Modified (intentional)
+
+| File | Reason |
+|------|--------|
+| `src/config/schema/oh-my-opencode-config.ts` | No change needed - `BackgroundTaskConfigSchema` is already composed into root schema via `background_task` field |
+| `src/create-managers.ts` | No change needed - `pluginConfig.background_task` already passed to `BackgroundManager` constructor |
+| `src/features/background-agent/manager.ts` | No change needed - already passes config to `ConcurrencyManager` |
+| `src/plugin-config.ts` | No change needed - `background_task` is a simple object field, uses default override merge |
+| `src/config/schema.ts` | No change needed - barrel already exports `BackgroundTaskConfigSchema` |
+
+## Design Decisions
+
+1. **Field name `maxBackgroundAgents`** - camelCase to match existing schema fields (`maxDepth`, `maxDescendants`, `defaultConcurrency`). The user-facing JSONC config key is also camelCase per existing convention in `background_task` section.
+
+2. **Global limit vs per-model limit** - The global limit is a ceiling across ALL concurrency keys. Per-model limits still apply independently. A task needs both a per-model slot AND a global slot to proceed.
+
+3. **Default of 5** - Matches the existing hardcoded default in `getConcurrencyLimit()`. When `maxBackgroundAgents` is not set, no global limit is enforced (only per-model limits apply), preserving backward compatibility.
+
+4. **Queue behavior** - When global limit is reached, tasks wait in the same FIFO queue mechanism. The global check happens inside `acquire()` before the per-model check.
+
+5. **0 means Infinity** - Following the existing pattern where `defaultConcurrency: 0` means unlimited, `maxBackgroundAgents: 0` would also mean no global limit.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/without_skill/outputs/pr-description.md

@@ -0,0 +1,50 @@
+# PR Description
+
+**Title:** feat: add `maxBackgroundAgents` config to limit total simultaneous background agents
+
+**Body:**
+
+## Summary
+
+- Add `maxBackgroundAgents` field to `BackgroundTaskConfigSchema` that enforces a global ceiling on total running background agents across all models/providers
+- Modify `ConcurrencyManager` to track global count and enforce the limit alongside existing per-model limits
+- Add schema validation tests and concurrency enforcement tests
+
+## Motivation
+
+Currently, concurrency is only limited per model/provider key (default 5 per key). On resource-constrained machines or when using many different models, the total number of background agents can grow unbounded (5 per model x N models). This config option lets users set a hard ceiling.
+
+## Changes
+
+### Schema (`src/config/schema/background-task.ts`)
+- Added `maxBackgroundAgents: z.number().int().min(1).optional()` to `BackgroundTaskConfigSchema`
+- Grouped with existing limit fields (`maxDepth`, `maxDescendants`)
+
+### ConcurrencyManager (`src/features/background-agent/concurrency.ts`)
+- Added `globalCount` tracking total active agents across all concurrency keys
+- Added `getGlobalLimit()` reading `maxBackgroundAgents` from config (defaults to `Infinity` = no global limit)
+- Modified `acquire()` to check both per-model AND global capacity
+- Modified `release()` to decrement global count and drain cross-model waiters blocked by global limit
+- Modified `clear()` to reset global state
+- Added `getGlobalCount()` / `getGlobalQueueLength()` for testing
+
+### Tests
+- `src/config/schema/background-task.test.ts`: 6 test cases for schema validation (valid, min boundary, below min, negative, non-integer, undefined)
+- `src/features/background-agent/concurrency.test.ts`: 8 test cases for global limit enforcement (cross-model blocking, release unblocking, per-model vs global interaction, no-config default, clear reset)
+
+## Config Example
+
+```jsonc
+{
+  "background_task": {
+    "maxBackgroundAgents": 5,
+    "defaultConcurrency": 3
+  }
+}
+```
+
+## Backward Compatibility
+
+- When `maxBackgroundAgents` is not set (default), no global limit is enforced - behavior is identical to before
+- Existing `defaultConcurrency`, `providerConcurrency`, and `modelConcurrency` continue to work unchanged
+- No config migration needed

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/without_skill/outputs/verification-strategy.md

@@ -0,0 +1,111 @@
+# Verification Strategy
+
+## 1. Static Analysis
+
+### TypeScript Typecheck
+```bash
+bun run typecheck
+```
+- Verify no type errors introduced
+- `BackgroundTaskConfig` type is inferred from Zod schema, so adding the field automatically updates the type
+- All existing consumers of `BackgroundTaskConfig` remain compatible (new field is optional)
+
+### LSP Diagnostics
+Check changed files for errors:
+- `src/config/schema/background-task.ts`
+- `src/features/background-agent/concurrency.ts`
+- `src/config/schema/background-task.test.ts`
+- `src/features/background-agent/concurrency.test.ts`
+
+## 2. Unit Tests
+
+### Schema Validation Tests
+```bash
+bun test src/config/schema/background-task.test.ts
+```
+
+| Test Case | Input | Expected |
+|-----------|-------|----------|
+| Valid value (10) | `{ maxBackgroundAgents: 10 }` | Parses to `10` |
+| Minimum boundary (1) | `{ maxBackgroundAgents: 1 }` | Parses to `1` |
+| Below minimum (0) | `{ maxBackgroundAgents: 0 }` | Throws `ZodError` |
+| Negative (-1) | `{ maxBackgroundAgents: -1 }` | Throws `ZodError` |
+| Non-integer (2.5) | `{ maxBackgroundAgents: 2.5 }` | Throws `ZodError` |
+| Not provided | `{}` | Field is `undefined` |
+
+### ConcurrencyManager Tests
+```bash
+bun test src/features/background-agent/concurrency.test.ts
+```
+
+| Test Case | Setup | Expected |
+|-----------|-------|----------|
+| No config = no global limit | No `maxBackgroundAgents` | `getGlobalLimit()` returns `Infinity` |
+| Config respected | `maxBackgroundAgents: 3` | `getGlobalLimit()` returns `3` |
+| Cross-model blocking | Global limit 2, acquire model-a + model-b, try model-c | model-c blocks |
+| Under-limit allows | Global limit 3, acquire 3 different models | All succeed |
+| Per-model + global interaction | Per-model 1, global 3, acquire model-a twice | Blocked by per-model, not global |
+| Release unblocks | Global limit 1, acquire model-a, queue model-b, release model-a | model-b proceeds |
+| No global limit = no enforcement | No config, acquire 6 different models | All succeed |
+| Clear resets global count | Acquire 2, clear | `getGlobalCount()` is 0 |
+
+### Existing Test Regression
+```bash
+bun test src/features/background-agent/concurrency.test.ts
+bun test src/config/schema/background-task.test.ts
+bun test src/config/schema.test.ts
+```
+All existing tests must continue to pass unchanged.
+
+## 3. Integration Verification
+
+### Config Loading Path
+Verify the config flows correctly through the system:
+
+1. **Schema → Type**: `BackgroundTaskConfig` type auto-includes `maxBackgroundAgents` via `z.infer`
+2. **Config file → Schema**: `loadConfigFromPath()` in `plugin-config.ts` uses `OhMyOpenCodeConfigSchema.safeParse()` which includes `BackgroundTaskConfigSchema`
+3. **Config → Manager**: `create-managers.ts` passes `pluginConfig.background_task` to `BackgroundManager` constructor
+4. **Manager → ConcurrencyManager**: `BackgroundManager` constructor passes config to `new ConcurrencyManager(config)`
+5. **ConcurrencyManager → Enforcement**: `acquire()` reads `config.maxBackgroundAgents` via `getGlobalLimit()`
+
+No changes needed in steps 2-4 since the field is optional and the existing plumbing passes the entire `BackgroundTaskConfig` object.
+
+### Manual Config Test
+Create a test config to verify parsing:
+```bash
+echo '{ "background_task": { "maxBackgroundAgents": 3 } }' | bun -e "
+  const { BackgroundTaskConfigSchema } = require('./src/config/schema/background-task');
+  const result = BackgroundTaskConfigSchema.safeParse(JSON.parse(require('fs').readFileSync('/dev/stdin', 'utf-8')).background_task);
+  console.log(result.success, result.data);
+"
+```
+
+## 4. Build Verification
+
+```bash
+bun run build
+```
+- Verify build succeeds
+- Schema JSON output includes the new field (if applicable)
+
+## 5. Edge Cases to Verify
+
+| Edge Case | Expected Behavior |
+|-----------|-------------------|
+| `maxBackgroundAgents` not set | No global limit enforced (backward compatible) |
+| `maxBackgroundAgents: 1` | Only 1 background agent at a time across all models |
+| `maxBackgroundAgents` > sum of all per-model limits | Global limit never triggers (per-model limits are tighter) |
+| Per-model limit tighter than global | Per-model limit blocks first |
+| Global limit tighter than per-model | Global limit blocks first |
+| Release from one model unblocks different model | Global slot freed, different model's waiter proceeds |
+| Manager shutdown with global waiters | `clear()` rejects all waiters and resets global count |
+| Concurrent acquire/release | No race conditions (single-threaded JS event loop) |
+
+## 6. CI Pipeline
+
+The existing CI workflow (`ci.yml`) will run:
+- `bun run typecheck` - type checking
+- `bun test` - all tests including new ones
+- `bun run build` - build verification
+
+No CI changes needed.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-1/without_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 365000, "total_duration_seconds": 365}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/eval_metadata.json

@@ -0,0 +1,37 @@
+{
+  "eval_id": 2,
+  "eval_name": "bugfix-atlas-null-check",
+  "prompt": "The atlas hook has a bug where it crashes when boulder.json is missing the worktree_path field. Fix it and land the fix as a PR. Make sure CI passes.",
+  "assertions": [
+    {
+      "id": "worktree-isolation",
+      "text": "Plan uses git worktree in a sibling directory",
+      "type": "manual"
+    },
+    {
+      "id": "minimal-fix",
+      "text": "Fix is minimal — adds null check, doesn't refactor unrelated code",
+      "type": "manual"
+    },
+    {
+      "id": "test-added",
+      "text": "Test case added for the missing worktree_path scenario",
+      "type": "manual"
+    },
+    {
+      "id": "three-gates",
+      "text": "Verification loop includes all 3 gates: CI, review-work, Cubic",
+      "type": "manual"
+    },
+    {
+      "id": "real-atlas-files",
+      "text": "References actual atlas hook files in src/hooks/atlas/",
+      "type": "manual"
+    },
+    {
+      "id": "fix-branch-naming",
+      "text": "Branch name follows fix/ prefix convention",
+      "type": "manual"
+    }
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/with_skill/grading.json

@@ -0,0 +1,11 @@
+{
+  "run_id": "eval-2-with_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": true, "evidence": "../omo-wt/fix-atlas-worktree-path-crash"},
+    {"text": "Fix is minimal — adds null check, doesn't refactor unrelated code", "passed": true, "evidence": "3 targeted changes: readBoulderState sanitization, idle-event guard, tests"},
+    {"text": "Test case added for the missing worktree_path scenario", "passed": true, "evidence": "Tests for missing and null worktree_path"},
+    {"text": "Verification loop includes all 3 gates", "passed": true, "evidence": "Gate A (CI), Gate B (review-work), Gate C (Cubic)"},
+    {"text": "References actual atlas hook files", "passed": true, "evidence": "src/hooks/atlas/idle-event.ts, src/features/boulder-state/storage.ts"},
+    {"text": "Branch name follows fix/ prefix convention", "passed": true, "evidence": "fix/atlas-worktree-path-crash"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/with_skill/outputs/code-changes.md

@@ -0,0 +1,205 @@
+# Code Changes
+
+## File 1: `src/features/boulder-state/storage.ts`
+
+**Change**: Add `worktree_path` sanitization in `readBoulderState()`
+
+```typescript
+// BEFORE (lines 29-32):
+    if (!Array.isArray(parsed.session_ids)) {
+      parsed.session_ids = []
+    }
+    return parsed as BoulderState
+
+// AFTER:
+    if (!Array.isArray(parsed.session_ids)) {
+      parsed.session_ids = []
+    }
+    if (parsed.worktree_path !== undefined && typeof parsed.worktree_path !== "string") {
+      parsed.worktree_path = undefined
+    }
+    return parsed as BoulderState
+```
+
+**Rationale**: `readBoulderState` casts raw `JSON.parse()` output as `BoulderState` without validating individual fields. When boulder.json has `"worktree_path": null` (valid JSON from manual edits, corrupted state, or external tools), the runtime type is `null` but TypeScript type says `string | undefined`. This sanitization ensures downstream code always gets the correct type.
+
+---
+
+## File 2: `src/hooks/atlas/idle-event.ts`
+
+**Change**: Add defensive string type guard before passing `worktree_path` to continuation functions.
+
+```typescript
+// BEFORE (lines 83-88 in scheduleRetry):
+      await injectContinuation({
+        ctx,
+        sessionID,
+        sessionState,
+        options,
+        planName: currentBoulder.plan_name,
+        progress: currentProgress,
+        agent: currentBoulder.agent,
+        worktreePath: currentBoulder.worktree_path,
+      })
+
+// AFTER:
+      await injectContinuation({
+        ctx,
+        sessionID,
+        sessionState,
+        options,
+        planName: currentBoulder.plan_name,
+        progress: currentProgress,
+        agent: currentBoulder.agent,
+        worktreePath: typeof currentBoulder.worktree_path === "string" ? currentBoulder.worktree_path : undefined,
+      })
+```
+
+```typescript
+// BEFORE (lines 184-188 in handleAtlasSessionIdle):
+  await injectContinuation({
+    ctx,
+    sessionID,
+    sessionState,
+    options,
+    planName: boulderState.plan_name,
+    progress,
+    agent: boulderState.agent,
+    worktreePath: boulderState.worktree_path,
+  })
+
+// AFTER:
+  await injectContinuation({
+    ctx,
+    sessionID,
+    sessionState,
+    options,
+    planName: boulderState.plan_name,
+    progress,
+    agent: boulderState.agent,
+    worktreePath: typeof boulderState.worktree_path === "string" ? boulderState.worktree_path : undefined,
+  })
+```
+
+**Rationale**: Belt-and-suspenders defense. Even though `readBoulderState` now sanitizes, direct `writeBoulderState` calls elsewhere could still produce invalid state. The `typeof` check is zero-cost and prevents any possibility of `null` or non-string values leaking through.
+
+---
+
+## File 3: `src/hooks/atlas/index.test.ts`
+
+**Change**: Add test cases for missing `worktree_path` scenarios within the existing `session.idle handler` describe block.
+
+```typescript
+    test("should inject continuation when boulder.json has no worktree_path field", async () => {
+      // given - boulder state WITHOUT worktree_path
+      const planPath = join(TEST_DIR, "test-plan.md")
+      writeFileSync(planPath, "# Plan\n- [ ] Task 1\n- [x] Task 2")
+
+      const state: BoulderState = {
+        active_plan: planPath,
+        started_at: "2026-01-02T10:00:00Z",
+        session_ids: [MAIN_SESSION_ID],
+        plan_name: "test-plan",
+      }
+      writeBoulderState(TEST_DIR, state)
+
+      const readState = readBoulderState(TEST_DIR)
+      expect(readState?.worktree_path).toBeUndefined()
+
+      const mockInput = createMockPluginInput()
+      const hook = createAtlasHook(mockInput)
+
+      // when
+      await hook.handler({
+        event: {
+          type: "session.idle",
+          properties: { sessionID: MAIN_SESSION_ID },
+        },
+      })
+
+      // then - continuation injected, no worktree context in prompt
+      expect(mockInput._promptMock).toHaveBeenCalled()
+      const callArgs = mockInput._promptMock.mock.calls[0][0]
+      expect(callArgs.body.parts[0].text).not.toContain("[Worktree:")
+      expect(callArgs.body.parts[0].text).toContain("1 remaining")
+    })
+
+    test("should handle boulder.json with worktree_path: null without crashing", async () => {
+      // given - manually write boulder.json with worktree_path: null (corrupted state)
+      const planPath = join(TEST_DIR, "test-plan.md")
+      writeFileSync(planPath, "# Plan\n- [ ] Task 1\n- [x] Task 2")
+
+      const boulderPath = join(SISYPHUS_DIR, "boulder.json")
+      writeFileSync(boulderPath, JSON.stringify({
+        active_plan: planPath,
+        started_at: "2026-01-02T10:00:00Z",
+        session_ids: [MAIN_SESSION_ID],
+        plan_name: "test-plan",
+        worktree_path: null,
+      }, null, 2))
+
+      const mockInput = createMockPluginInput()
+      const hook = createAtlasHook(mockInput)
+
+      // when
+      await hook.handler({
+        event: {
+          type: "session.idle",
+          properties: { sessionID: MAIN_SESSION_ID },
+        },
+      })
+
+      // then - should inject continuation without crash, no "[Worktree: null]"
+      expect(mockInput._promptMock).toHaveBeenCalled()
+      const callArgs = mockInput._promptMock.mock.calls[0][0]
+      expect(callArgs.body.parts[0].text).not.toContain("[Worktree: null]")
+      expect(callArgs.body.parts[0].text).not.toContain("[Worktree: undefined]")
+    })
+```
+
+---
+
+## File 4: `src/features/boulder-state/storage.test.ts` (addition to existing)
+
+**Change**: Add `readBoulderState` sanitization test.
+
+```typescript
+  describe("#given boulder.json with worktree_path: null", () => {
+    test("#then readBoulderState should sanitize null to undefined", () => {
+      // given
+      const boulderPath = join(TEST_DIR, ".sisyphus", "boulder.json")
+      writeFileSync(boulderPath, JSON.stringify({
+        active_plan: "/path/to/plan.md",
+        started_at: "2026-01-02T10:00:00Z",
+        session_ids: ["session-1"],
+        plan_name: "test-plan",
+        worktree_path: null,
+      }, null, 2))
+
+      // when
+      const state = readBoulderState(TEST_DIR)
+
+      // then
+      expect(state).not.toBeNull()
+      expect(state!.worktree_path).toBeUndefined()
+    })
+
+    test("#then readBoulderState should preserve valid worktree_path string", () => {
+      // given
+      const boulderPath = join(TEST_DIR, ".sisyphus", "boulder.json")
+      writeFileSync(boulderPath, JSON.stringify({
+        active_plan: "/path/to/plan.md",
+        started_at: "2026-01-02T10:00:00Z",
+        session_ids: ["session-1"],
+        plan_name: "test-plan",
+        worktree_path: "/valid/worktree/path",
+      }, null, 2))
+
+      // when
+      const state = readBoulderState(TEST_DIR)
+
+      // then
+      expect(state?.worktree_path).toBe("/valid/worktree/path")
+    })
+  })
+```

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/with_skill/outputs/execution-plan.md

@@ -0,0 +1,78 @@
+# Execution Plan — Fix atlas hook crash on missing worktree_path
+
+## Phase 0: Setup
+
+1. **Create worktree from origin/dev**:
+   ```bash
+   git fetch origin dev
+   git worktree add ../omo-wt/fix-atlas-worktree-path-crash origin/dev
+   ```
+2. **Create feature branch**:
+   ```bash
+   cd ../omo-wt/fix-atlas-worktree-path-crash
+   git checkout -b fix/atlas-worktree-path-crash
+   ```
+
+## Phase 1: Implement
+
+### Step 1: Fix `readBoulderState()` in `src/features/boulder-state/storage.ts`
+- Add `worktree_path` sanitization after JSON parse
+- Ensure `worktree_path` is `string | undefined`, never `null` or other types
+- This is the root cause: raw `JSON.parse` + `as BoulderState` cast allows type violations at runtime
+
+### Step 2: Add defensive guard in `src/hooks/atlas/idle-event.ts`
+- Before passing `boulderState.worktree_path` to `injectContinuation`, validate it's a string
+- Apply same guard in the `scheduleRetry` callback (line 86)
+- Ensures even if `readBoulderState` is bypassed, the idle handler won't crash
+
+### Step 3: Add test coverage in `src/hooks/atlas/index.test.ts`
+- Add test: boulder.json without `worktree_path` field → session.idle works
+- Add test: boulder.json with `worktree_path: null` → session.idle works (no `[Worktree: null]` in prompt)
+- Add test: `readBoulderState` sanitizes `null` worktree_path to `undefined`
+- Follow existing given/when/then test pattern
+
+### Step 4: Local validation
+```bash
+bun run typecheck
+bun test src/hooks/atlas/
+bun test src/features/boulder-state/
+bun run build
+```
+
+### Step 5: Atomic commit
+```bash
+git add src/features/boulder-state/storage.ts src/hooks/atlas/idle-event.ts src/hooks/atlas/index.test.ts
+git commit -m "fix(atlas): prevent crash when boulder.json missing worktree_path field
+
+readBoulderState() performs unsafe cast of parsed JSON as BoulderState.
+When worktree_path is absent or null in boulder.json, downstream code
+in idle-event.ts could receive null where string|undefined is expected.
+
+- Sanitize worktree_path in readBoulderState (reject non-string values)
+- Add defensive typeof check in idle-event before passing to continuation
+- Add test coverage for missing and null worktree_path scenarios"
+```
+
+## Phase 2: PR Creation
+
+```bash
+git push -u origin fix/atlas-worktree-path-crash
+gh pr create \
+  --base dev \
+  --title "fix(atlas): prevent crash when boulder.json missing worktree_path" \
+  --body-file /tmp/pull-request-atlas-worktree-fix.md
+```
+
+## Phase 3: Verify Loop
+
+- **Gate A (CI)**: `gh pr checks --watch` — wait for all checks green
+- **Gate B (review-work)**: Run 5-agent review (Oracle goal, Oracle quality, Oracle security, QA execution, context mining)
+- **Gate C (Cubic)**: Wait for cubic-dev-ai[bot] to respond "No issues found"
+- On any failure: fix-commit-push, re-enter verify loop
+
+## Phase 4: Merge
+
+```bash
+gh pr merge --squash --delete-branch
+git worktree remove ../omo-wt/fix-atlas-worktree-path-crash
+```

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/with_skill/outputs/pr-description.md

@@ -0,0 +1,42 @@
+# PR Title
+
+```
+fix(atlas): prevent crash when boulder.json missing worktree_path
+```
+
+# PR Body
+
+## Summary
+
+- Fix runtime type violation in atlas hook when `boulder.json` lacks `worktree_path` field
+- Add `worktree_path` sanitization in `readBoulderState()` to reject non-string values (e.g., `null` from manual edits)
+- Add defensive `typeof` guards in `idle-event.ts` before passing worktree path to continuation injection
+- Add test coverage for missing and null `worktree_path` scenarios
+
+## Problem
+
+`readBoulderState()` in `src/features/boulder-state/storage.ts` casts raw `JSON.parse()` output directly as `BoulderState` via `return parsed as BoulderState`. This bypasses TypeScript's type system entirely at runtime.
+
+When `boulder.json` is missing the `worktree_path` field (common for boulders created before worktree support was added, or created without `--worktree` flag), `boulderState.worktree_path` is `undefined` which is handled correctly. However, when boulder.json has `"worktree_path": null` (possible from manual edits, external tooling, or corrupted state), the runtime type becomes `null` which violates the TypeScript type `string | undefined`.
+
+This `null` value propagates through:
+1. `idle-event.ts:handleAtlasSessionIdle()` → `injectContinuation()` → `injectBoulderContinuation()`
+2. `idle-event.ts:scheduleRetry()` callback → same chain
+
+While the `boulder-continuation-injector.ts` handles falsy values via `worktreePath ? ... : ""`, the type mismatch can cause subtle downstream issues and violates the contract of the `BoulderState` interface.
+
+## Changes
+
+| File | Change |
+|------|--------|
+| `src/features/boulder-state/storage.ts` | Sanitize `worktree_path` in `readBoulderState()` — reject non-string values |
+| `src/hooks/atlas/idle-event.ts` | Add `typeof` guards before passing worktree_path to continuation (2 call sites) |
+| `src/hooks/atlas/index.test.ts` | Add 2 tests: missing worktree_path + null worktree_path in session.idle |
+| `src/features/boulder-state/storage.test.ts` | Add 2 tests: sanitization of null + preservation of valid string |
+
+## Testing
+
+- `bun test src/hooks/atlas/` — all existing + new tests pass
+- `bun test src/features/boulder-state/` — all existing + new tests pass
+- `bun run typecheck` — clean
+- `bun run build` — clean

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/with_skill/outputs/verification-strategy.md

@@ -0,0 +1,87 @@
+# Verification Strategy
+
+## Gate A: CI (`gh pr checks --watch`)
+
+### What CI runs (from `ci.yml`)
+1. **Tests (split)**: Mock-heavy tests in isolation + batch tests
+2. **Typecheck**: `bun run typecheck` (tsc --noEmit)
+3. **Build**: `bun run build` (ESM + declarations + schema)
+
+### Pre-push local validation
+Before pushing, run the exact CI steps locally to catch failures early:
+
+```bash
+# Targeted test runs first (fast feedback)
+bun test src/features/boulder-state/storage.test.ts
+bun test src/hooks/atlas/index.test.ts
+
+# Full test suite
+bun test
+
+# Type check
+bun run typecheck
+
+# Build
+bun run build
+```
+
+### Failure handling
+- **Test failure**: Read test output, fix code, create new commit (never amend pushed commits), push
+- **Typecheck failure**: Run `lsp_diagnostics` on changed files, fix type errors, commit, push
+- **Build failure**: Check build output for missing exports or circular deps, fix, commit, push
+
+After each fix-commit-push: `gh pr checks --watch` to re-enter gate
+
+## Gate B: review-work (5-agent review)
+
+### The 5 parallel agents
+1. **Oracle (goal/constraint verification)**: Checks the fix matches the stated problem — `worktree_path` crash resolved, no scope creep
+2. **Oracle (code quality)**: Validates code follows existing patterns — factory pattern, given/when/then tests, < 200 LOC, no catch-all files
+3. **Oracle (security)**: Ensures no new security issues — JSON parse injection, path traversal in worktree_path
+4. **QA agent (hands-on execution)**: Actually runs the tests, checks `lsp_diagnostics` on changed files, verifies the fix in action
+5. **Context mining agent**: Checks GitHub issues, git history, related PRs for context alignment
+
+### Expected focus areas for this PR
+- Oracle (goal): Does the sanitization in `readBoulderState` actually prevent the crash? Is the `typeof` guard necessary or redundant?
+- Oracle (quality): Are the new tests following the given/when/then pattern? Do they use the same mock setup as existing tests?
+- Oracle (security): Is the `worktree_path` value ever used in path operations without sanitization? (Answer: no, it's only used in template strings)
+- QA: Run `bun test src/hooks/atlas/index.test.ts` — does the null worktree_path test actually trigger the bug before fix?
+
+### Failure handling
+- Each oracle produces a PASS/FAIL verdict with specific issues
+- On FAIL: read the specific issue, fix in the worktree, commit, push, re-run review-work
+- All 5 agents must PASS
+
+## Gate C: Cubic (`cubic-dev-ai[bot]`)
+
+### What Cubic checks
+- Automated code review bot that analyzes the PR diff
+- Looks for: type safety issues, missing error handling, test coverage gaps, anti-patterns
+
+### Expected result
+- "No issues found" for this small, focused fix
+- 3 files changed (storage.ts, idle-event.ts, index.test.ts) + 1 test file
+
+### Failure handling
+- If Cubic flags an issue: evaluate if it's a real concern or false positive
+- Real concern: fix, commit, push
+- False positive: comment explaining why the flagged pattern is intentional
+- Wait for Cubic to re-review after push
+
+## Post-verification: Merge
+
+Once all 3 gates pass:
+```bash
+gh pr merge --squash --delete-branch
+git worktree remove ../omo-wt/fix-atlas-worktree-path-crash
+```
+
+On merge failure (conflicts):
+```bash
+cd ../omo-wt/fix-atlas-worktree-path-crash
+git fetch origin dev
+git rebase origin/dev
+# Resolve conflicts if any
+git push --force-with-lease
+# Re-enter verify loop from Gate A
+```

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/with_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 506000, "total_duration_seconds": 506}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/without_skill/grading.json

@@ -0,0 +1,11 @@
+{
+  "run_id": "eval-2-without_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": false, "evidence": "No worktree. Steps go directly to creating branch and modifying files."},
+    {"text": "Fix is minimal — adds null check, doesn't refactor unrelated code", "passed": true, "evidence": "Focused fix though also adds try/catch in setTimeout (reasonable secondary fix)"},
+    {"text": "Test case added for the missing worktree_path scenario", "passed": true, "evidence": "Detailed test plan for missing/null/malformed boulder.json"},
+    {"text": "Verification loop includes all 3 gates", "passed": false, "evidence": "Only mentions CI pipeline (step 5). No review-work or Cubic."},
+    {"text": "References actual atlas hook files", "passed": true, "evidence": "References idle-event.ts, storage.ts with line numbers"},
+    {"text": "Branch name follows fix/ prefix convention", "passed": true, "evidence": "fix/atlas-hook-missing-worktree-path"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/without_skill/outputs/code-changes.md

@@ -0,0 +1,334 @@
+# Code Changes: Fix Atlas Hook Crash on Missing worktree_path
+
+## Change 1: Harden `readBoulderState()` validation
+
+**File:** `src/features/boulder-state/storage.ts`
+
+### Before (lines 16-36):
+```typescript
+export function readBoulderState(directory: string): BoulderState | null {
+  const filePath = getBoulderFilePath(directory)
+
+  if (!existsSync(filePath)) {
+    return null
+  }
+
+  try {
+    const content = readFileSync(filePath, "utf-8")
+    const parsed = JSON.parse(content)
+    if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) {
+      return null
+    }
+    if (!Array.isArray(parsed.session_ids)) {
+      parsed.session_ids = []
+    }
+    return parsed as BoulderState
+  } catch {
+    return null
+  }
+}
+```
+
+### After:
+```typescript
+export function readBoulderState(directory: string): BoulderState | null {
+  const filePath = getBoulderFilePath(directory)
+
+  if (!existsSync(filePath)) {
+    return null
+  }
+
+  try {
+    const content = readFileSync(filePath, "utf-8")
+    const parsed = JSON.parse(content)
+    if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) {
+      return null
+    }
+    if (typeof parsed.active_plan !== "string" || typeof parsed.plan_name !== "string") {
+      return null
+    }
+    if (!Array.isArray(parsed.session_ids)) {
+      parsed.session_ids = []
+    }
+    if (parsed.worktree_path !== undefined && typeof parsed.worktree_path !== "string") {
+      delete parsed.worktree_path
+    }
+    return parsed as BoulderState
+  } catch {
+    return null
+  }
+}
+```
+
+**Rationale:** Validates that required fields (`active_plan`, `plan_name`) are strings. Strips `worktree_path` if it's present but not a string (e.g., `null`, number). This prevents downstream crashes from `existsSync(undefined)` and ensures type safety at the boundary.
+
+---
+
+## Change 2: Add try/catch in setTimeout retry callback
+
+**File:** `src/hooks/atlas/idle-event.ts`
+
+### Before (lines 62-88):
+```typescript
+sessionState.pendingRetryTimer = setTimeout(async () => {
+    sessionState.pendingRetryTimer = undefined
+
+    if (sessionState.promptFailureCount >= 2) return
+    if (sessionState.waitingForFinalWaveApproval) return
+
+    const currentBoulder = readBoulderState(ctx.directory)
+    if (!currentBoulder) return
+    if (!currentBoulder.session_ids?.includes(sessionID)) return
+
+    const currentProgress = getPlanProgress(currentBoulder.active_plan)
+    if (currentProgress.isComplete) return
+    if (options?.isContinuationStopped?.(sessionID)) return
+    if (options?.shouldSkipContinuation?.(sessionID)) return
+    if (hasRunningBackgroundTasks(sessionID, options)) return
+
+    await injectContinuation({
+      ctx,
+      sessionID,
+      sessionState,
+      options,
+      planName: currentBoulder.plan_name,
+      progress: currentProgress,
+      agent: currentBoulder.agent,
+      worktreePath: currentBoulder.worktree_path,
+    })
+  }, RETRY_DELAY_MS)
+```
+
+### After:
+```typescript
+sessionState.pendingRetryTimer = setTimeout(async () => {
+    sessionState.pendingRetryTimer = undefined
+
+    try {
+      if (sessionState.promptFailureCount >= 2) return
+      if (sessionState.waitingForFinalWaveApproval) return
+
+      const currentBoulder = readBoulderState(ctx.directory)
+      if (!currentBoulder) return
+      if (!currentBoulder.session_ids?.includes(sessionID)) return
+
+      const currentProgress = getPlanProgress(currentBoulder.active_plan)
+      if (currentProgress.isComplete) return
+      if (options?.isContinuationStopped?.(sessionID)) return
+      if (options?.shouldSkipContinuation?.(sessionID)) return
+      if (hasRunningBackgroundTasks(sessionID, options)) return
+
+      await injectContinuation({
+        ctx,
+        sessionID,
+        sessionState,
+        options,
+        planName: currentBoulder.plan_name,
+        progress: currentProgress,
+        agent: currentBoulder.agent,
+        worktreePath: currentBoulder.worktree_path,
+      })
+    } catch (error) {
+      log(`[${HOOK_NAME}] Retry continuation failed`, { sessionID, error: String(error) })
+    }
+  }, RETRY_DELAY_MS)
+```
+
+**Rationale:** The async callback in setTimeout creates a floating promise. Without try/catch, any error becomes an unhandled rejection that can crash the process. This is the critical safety net even after the `readBoulderState` fix.
+
+---
+
+## Change 3: Defensive guard in `getPlanProgress`
+
+**File:** `src/features/boulder-state/storage.ts`
+
+### Before (lines 115-118):
+```typescript
+export function getPlanProgress(planPath: string): PlanProgress {
+  if (!existsSync(planPath)) {
+    return { total: 0, completed: 0, isComplete: true }
+  }
+```
+
+### After:
+```typescript
+export function getPlanProgress(planPath: string): PlanProgress {
+  if (typeof planPath !== "string" || !existsSync(planPath)) {
+    return { total: 0, completed: 0, isComplete: true }
+  }
+```
+
+**Rationale:** Defense-in-depth. Even though `readBoulderState` now validates `active_plan`, the `getPlanProgress` function is a public API that could be called from other paths with invalid input. A `typeof` check before `existsSync` prevents the TypeError from `existsSync(undefined)`.
+
+---
+
+## Change 4: New tests
+
+### File: `src/features/boulder-state/storage.test.ts` (additions)
+
+```typescript
+test("should return null when active_plan is missing", () => {
+  // given - boulder.json without active_plan
+  const boulderFile = join(SISYPHUS_DIR, "boulder.json")
+  writeFileSync(boulderFile, JSON.stringify({
+    started_at: "2026-01-01T00:00:00Z",
+    session_ids: ["ses-1"],
+    plan_name: "plan",
+  }))
+
+  // when
+  const result = readBoulderState(TEST_DIR)
+
+  // then
+  expect(result).toBeNull()
+})
+
+test("should return null when plan_name is missing", () => {
+  // given - boulder.json without plan_name
+  const boulderFile = join(SISYPHUS_DIR, "boulder.json")
+  writeFileSync(boulderFile, JSON.stringify({
+    active_plan: "/path/to/plan.md",
+    started_at: "2026-01-01T00:00:00Z",
+    session_ids: ["ses-1"],
+  }))
+
+  // when
+  const result = readBoulderState(TEST_DIR)
+
+  // then
+  expect(result).toBeNull()
+})
+
+test("should strip non-string worktree_path from boulder state", () => {
+  // given - boulder.json with worktree_path set to null
+  const boulderFile = join(SISYPHUS_DIR, "boulder.json")
+  writeFileSync(boulderFile, JSON.stringify({
+    active_plan: "/path/to/plan.md",
+    started_at: "2026-01-01T00:00:00Z",
+    session_ids: ["ses-1"],
+    plan_name: "plan",
+    worktree_path: null,
+  }))
+
+  // when
+  const result = readBoulderState(TEST_DIR)
+
+  // then
+  expect(result).not.toBeNull()
+  expect(result!.worktree_path).toBeUndefined()
+})
+
+test("should preserve valid worktree_path string", () => {
+  // given - boulder.json with valid worktree_path
+  const boulderFile = join(SISYPHUS_DIR, "boulder.json")
+  writeFileSync(boulderFile, JSON.stringify({
+    active_plan: "/path/to/plan.md",
+    started_at: "2026-01-01T00:00:00Z",
+    session_ids: ["ses-1"],
+    plan_name: "plan",
+    worktree_path: "/valid/worktree/path",
+  }))
+
+  // when
+  const result = readBoulderState(TEST_DIR)
+
+  // then
+  expect(result).not.toBeNull()
+  expect(result!.worktree_path).toBe("/valid/worktree/path")
+})
+```
+
+### File: `src/features/boulder-state/storage.test.ts` (getPlanProgress additions)
+
+```typescript
+test("should handle undefined planPath without crashing", () => {
+  // given - undefined as planPath (from malformed boulder state)
+
+  // when
+  const progress = getPlanProgress(undefined as unknown as string)
+
+  // then
+  expect(progress.total).toBe(0)
+  expect(progress.isComplete).toBe(true)
+})
+```
+
+### File: `src/hooks/atlas/index.test.ts` (additions to session.idle section)
+
+```typescript
+test("should handle boulder state without worktree_path gracefully", async () => {
+  // given - boulder state with incomplete plan, no worktree_path
+  const planPath = join(TEST_DIR, "test-plan.md")
+  writeFileSync(planPath, "# Plan\n- [ ] Task 1\n- [x] Task 2")
+
+  const state: BoulderState = {
+    active_plan: planPath,
+    started_at: "2026-01-02T10:00:00Z",
+    session_ids: [MAIN_SESSION_ID],
+    plan_name: "test-plan",
+    // worktree_path intentionally omitted
+  }
+  writeBoulderState(TEST_DIR, state)
+
+  const mockInput = createMockPluginInput()
+  const hook = createAtlasHook(mockInput)
+
+  // when
+  await hook.handler({
+    event: {
+      type: "session.idle",
+      properties: { sessionID: MAIN_SESSION_ID },
+    },
+  })
+
+  // then - should call prompt without crashing, continuation should not contain worktree context
+  expect(mockInput._promptMock).toHaveBeenCalled()
+  const callArgs = mockInput._promptMock.mock.calls[0][0]
+  expect(callArgs.body.parts[0].text).toContain("incomplete tasks")
+  expect(callArgs.body.parts[0].text).not.toContain("[Worktree:")
+})
+
+test("should include worktree context when worktree_path is present in boulder state", async () => {
+  // given - boulder state with worktree_path
+  const planPath = join(TEST_DIR, "test-plan.md")
+  writeFileSync(planPath, "# Plan\n- [ ] Task 1")
+
+  const state: BoulderState = {
+    active_plan: planPath,
+    started_at: "2026-01-02T10:00:00Z",
+    session_ids: [MAIN_SESSION_ID],
+    plan_name: "test-plan",
+    worktree_path: "/some/worktree/path",
+  }
+  writeBoulderState(TEST_DIR, state)
+
+  const mockInput = createMockPluginInput()
+  const hook = createAtlasHook(mockInput)
+
+  // when
+  await hook.handler({
+    event: {
+      type: "session.idle",
+      properties: { sessionID: MAIN_SESSION_ID },
+    },
+  })
+
+  // then - should include worktree context in continuation prompt
+  expect(mockInput._promptMock).toHaveBeenCalled()
+  const callArgs = mockInput._promptMock.mock.calls[0][0]
+  expect(callArgs.body.parts[0].text).toContain("[Worktree: /some/worktree/path]")
+})
+```
+
+---
+
+## Summary of Changes
+
+| File | Change | Lines Modified |
+|------|--------|---------------|
+| `src/features/boulder-state/storage.ts` | Validate required fields + sanitize worktree_path + guard getPlanProgress | ~8 lines added |
+| `src/hooks/atlas/idle-event.ts` | try/catch around setTimeout async callback | ~4 lines added |
+| `src/features/boulder-state/storage.test.ts` | 5 new tests for validation | ~60 lines added |
+| `src/hooks/atlas/index.test.ts` | 2 new tests for worktree_path handling | ~50 lines added |
+
+Total: ~4 production lines changed, ~8 defensive lines added, ~110 test lines added.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/without_skill/outputs/execution-plan.md

@@ -0,0 +1,86 @@
+# Execution Plan: Fix Atlas Hook Crash on Missing worktree_path
+
+## Bug Analysis
+
+### Root Cause
+
+`readBoulderState()` in `src/features/boulder-state/storage.ts` performs minimal validation when parsing `boulder.json`:
+
+```typescript
+const parsed = JSON.parse(content)
+if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) return null
+if (!Array.isArray(parsed.session_ids)) parsed.session_ids = []
+return parsed as BoulderState  // <-- unsafe cast, no field validation
+```
+
+It validates `session_ids` but NOT `active_plan`, `plan_name`, or `worktree_path`. This means a malformed `boulder.json` (e.g., `{}` or missing key fields) passes through and downstream code crashes.
+
+### Crash Path
+
+1. `boulder.json` is written without required fields (manual edit, corruption, partial write)
+2. `readBoulderState()` returns it as `BoulderState` with `active_plan: undefined`
+3. Multiple call sites pass `boulderState.active_plan` to `getPlanProgress(planPath: string)`:
+   - `src/hooks/atlas/idle-event.ts:72` (inside `setTimeout` callback - unhandled rejection!)
+   - `src/hooks/atlas/resolve-active-boulder-session.ts:21`
+   - `src/hooks/atlas/tool-execute-after.ts:74`
+4. `getPlanProgress()` calls `existsSync(undefined)` which throws: `TypeError: The "path" argument must be of type string`
+
+### worktree_path-Specific Issues
+
+When `worktree_path` field is missing from `boulder.json`:
+- The `idle-event.ts` `scheduleRetry` setTimeout callback (lines 62-88) has NO try/catch. An unhandled promise rejection from the async callback crashes the process.
+- `readBoulderState()` returns `worktree_path: undefined` which itself is handled in `boulder-continuation-injector.ts` (line 42 uses truthiness check), but the surrounding code in the setTimeout lacks error protection.
+
+### Secondary Issue: Unhandled Promise in setTimeout
+
+In `idle-event.ts` lines 62-88:
+```typescript
+sessionState.pendingRetryTimer = setTimeout(async () => {
+  // ... no try/catch wrapper
+  const currentBoulder = readBoulderState(ctx.directory)
+  const currentProgress = getPlanProgress(currentBoulder.active_plan)  // CRASH if active_plan undefined
+  // ...
+}, RETRY_DELAY_MS)
+```
+
+The async callback creates a floating promise. Any thrown error becomes an unhandled rejection.
+
+---
+
+## Step-by-Step Plan
+
+### Step 1: Harden `readBoulderState()` validation
+**File:** `src/features/boulder-state/storage.ts`
+
+- After the `session_ids` fix, add validation for `active_plan` and `plan_name` (required fields)
+- Validate `worktree_path` is either `undefined` or a string (not `null`, not a number)
+- Return `null` for boulder states with missing required fields
+
+### Step 2: Add try/catch in setTimeout callback
+**File:** `src/hooks/atlas/idle-event.ts`
+
+- Wrap the `setTimeout` async callback body in try/catch
+- Log errors with the atlas hook logger
+
+### Step 3: Add defensive guard in `getPlanProgress`
+**File:** `src/features/boulder-state/storage.ts`
+
+- Add early return for non-string `planPath` argument
+
+### Step 4: Add tests
+**Files:**
+- `src/features/boulder-state/storage.test.ts` - test missing/malformed fields
+- `src/hooks/atlas/index.test.ts` - test atlas hook with boulder missing worktree_path
+
+### Step 5: Run CI checks
+```bash
+bun run typecheck
+bun test src/features/boulder-state/storage.test.ts
+bun test src/hooks/atlas/index.test.ts
+bun test  # full suite
+```
+
+### Step 6: Create PR
+- Branch: `fix/atlas-hook-missing-worktree-path`
+- Target: `dev`
+- Run CI and verify passes

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/without_skill/outputs/pr-description.md

@@ -0,0 +1,23 @@
+## Summary
+
+- Fix crash in atlas hook when `boulder.json` is missing `worktree_path` (or other required fields) by hardening `readBoulderState()` validation
+- Wrap the unprotected `setTimeout` retry callback in `idle-event.ts` with try/catch to prevent unhandled promise rejections
+- Add defensive type guard in `getPlanProgress()` to prevent `existsSync(undefined)` TypeError
+
+## Context
+
+When `boulder.json` is malformed or manually edited to omit fields, `readBoulderState()` returns an object cast as `BoulderState` without validating required fields. Downstream callers like `getPlanProgress(boulderState.active_plan)` then pass `undefined` to `existsSync()`, which throws a TypeError. This crash is especially dangerous in the `setTimeout` retry callback in `idle-event.ts`, where the error becomes an unhandled promise rejection.
+
+## Changes
+
+### `src/features/boulder-state/storage.ts`
+- `readBoulderState()`: Validate `active_plan` and `plan_name` are strings (return `null` if not)
+- `readBoulderState()`: Strip `worktree_path` if present but not a string type
+- `getPlanProgress()`: Add `typeof planPath !== "string"` guard before `existsSync`
+
+### `src/hooks/atlas/idle-event.ts`
+- Wrap `scheduleRetry` setTimeout async callback body in try/catch
+
+### Tests
+- `src/features/boulder-state/storage.test.ts`: 5 new tests for missing/malformed fields
+- `src/hooks/atlas/index.test.ts`: 2 new tests for worktree_path presence/absence in continuation prompt

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/without_skill/outputs/verification-strategy.md

@@ -0,0 +1,119 @@
+# Verification Strategy
+
+## 1. Unit Tests (Direct Verification)
+
+### boulder-state storage tests
+```bash
+bun test src/features/boulder-state/storage.test.ts
+```
+
+Verify:
+- `readBoulderState()` returns `null` when `active_plan` missing
+- `readBoulderState()` returns `null` when `plan_name` missing
+- `readBoulderState()` strips non-string `worktree_path` (e.g., `null`)
+- `readBoulderState()` preserves valid string `worktree_path`
+- `getPlanProgress(undefined)` returns safe default without crashing
+- Existing tests still pass (session_ids defaults, empty object, etc.)
+
+### atlas hook tests
+```bash
+bun test src/hooks/atlas/index.test.ts
+```
+
+Verify:
+- session.idle handler works with boulder state missing `worktree_path` (no crash, prompt injected)
+- session.idle handler includes `[Worktree: ...]` context when `worktree_path` IS present
+- All 30+ existing tests still pass
+
+### atlas idle-event lineage tests
+```bash
+bun test src/hooks/atlas/idle-event-lineage.test.ts
+```
+
+Verify existing lineage tests unaffected.
+
+### start-work hook tests
+```bash
+bun test src/hooks/start-work/index.test.ts
+```
+
+Verify worktree-related start-work tests still pass (these create boulder states with/without `worktree_path`).
+
+## 2. Type Safety
+
+```bash
+bun run typecheck
+```
+
+Verify zero new TypeScript errors. The changes are purely additive runtime guards that align with existing types (`worktree_path?: string`).
+
+## 3. LSP Diagnostics on Changed Files
+
+```
+lsp_diagnostics on:
+  - src/features/boulder-state/storage.ts
+  - src/hooks/atlas/idle-event.ts
+```
+
+Verify zero errors/warnings.
+
+## 4. Full Test Suite
+
+```bash
+bun test
+```
+
+Verify no regressions across the entire codebase.
+
+## 5. Build
+
+```bash
+bun run build
+```
+
+Verify build succeeds.
+
+## 6. Manual Smoke Test (Reproduction)
+
+To manually verify the fix:
+
+```bash
+# Create a malformed boulder.json (missing worktree_path)
+mkdir -p .sisyphus
+echo '{"active_plan": ".sisyphus/plans/test.md", "plan_name": "test", "session_ids": ["ses-1"]}' > .sisyphus/boulder.json
+
+# Create a plan file
+mkdir -p .sisyphus/plans
+echo '# Plan\n- [ ] Task 1' > .sisyphus/plans/test.md
+
+# Start opencode - atlas hook should NOT crash when session.idle fires
+# Verify /tmp/oh-my-opencode.log shows normal continuation behavior
+```
+
+Also test the extreme case:
+```bash
+# boulder.json with no required fields
+echo '{}' > .sisyphus/boulder.json
+
+# After fix: readBoulderState returns null, atlas hook gracefully skips
+```
+
+## 7. CI Pipeline
+
+After pushing the branch, verify:
+- `ci.yml` workflow passes: tests (split: mock-heavy isolated + batch), typecheck, build
+- No new lint warnings
+
+## 8. Edge Cases Covered
+
+| Scenario | Expected Behavior |
+|----------|-------------------|
+| `boulder.json` = `{}` | `readBoulderState` returns `null` |
+| `boulder.json` missing `active_plan` | `readBoulderState` returns `null` |
+| `boulder.json` missing `plan_name` | `readBoulderState` returns `null` |
+| `boulder.json` has `worktree_path: null` | Field stripped, returned as `undefined` |
+| `boulder.json` has `worktree_path: 42` | Field stripped, returned as `undefined` |
+| `boulder.json` has no `worktree_path` | Works normally, no crash |
+| `boulder.json` has valid `worktree_path` | Preserved, included in continuation prompt |
+| setTimeout retry with corrupted boulder.json | Error caught and logged, no process crash |
+| `getPlanProgress(undefined)` | Returns `{ total: 0, completed: 0, isComplete: true }` |

.opencode/skills/work-with-pr-workspace/iteration-1/eval-2/without_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 325000, "total_duration_seconds": 325}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/eval_metadata.json

@@ -0,0 +1,32 @@
+{
+  "eval_id": 3,
+  "eval_name": "refactor-split-constants",
+  "prompt": "Refactor src/tools/delegate-task/constants.ts to split DEFAULT_CATEGORIES and CATEGORY_MODEL_REQUIREMENTS into separate files. Keep backward compatibility with the barrel export. Make a PR.",
+  "assertions": [
+    {
+      "id": "worktree-isolation",
+      "text": "Plan uses git worktree in a sibling directory",
+      "type": "manual"
+    },
+    {
+      "id": "multiple-atomic-commits",
+      "text": "Uses 2+ commits for the multi-file refactor",
+      "type": "manual"
+    },
+    {
+      "id": "barrel-export",
+      "text": "Maintains backward compatibility via barrel re-export in constants.ts or index.ts",
+      "type": "manual"
+    },
+    {
+      "id": "three-gates",
+      "text": "Verification loop includes all 3 gates",
+      "type": "manual"
+    },
+    {
+      "id": "real-constants-file",
+      "text": "References actual src/tools/delegate-task/constants.ts file and its exports",
+      "type": "manual"
+    }
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/with_skill/grading.json

@@ -0,0 +1,10 @@
+{
+  "run_id": "eval-3-with_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": true, "evidence": "../omo-wt/refactor-delegate-task-constants"},
+    {"text": "Uses 2+ commits for the multi-file refactor", "passed": true, "evidence": "Commit 1: category defaults+appends, Commit 2: plan agent prompt+names"},
+    {"text": "Maintains backward compatibility via barrel re-export", "passed": true, "evidence": "constants.ts converted to re-export from 4 new files, full import map verified"},
+    {"text": "Verification loop includes all 3 gates", "passed": true, "evidence": "Gate A (CI), Gate B (review-work), Gate C (Cubic)"},
+    {"text": "References actual src/tools/delegate-task/constants.ts", "passed": true, "evidence": "654 lines analyzed, 4 responsibilities identified, full external+internal import map"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/with_skill/outputs/code-changes.md

@@ -0,0 +1,221 @@
+# Code Changes
+
+## New File: `src/tools/delegate-task/default-categories.ts`
+
+```typescript
+import type { CategoryConfig } from "../../config/schema"
+
+export const DEFAULT_CATEGORIES: Record<string, CategoryConfig> = {
+  "visual-engineering": { model: "google/gemini-3.1-pro", variant: "high" },
+  ultrabrain: { model: "openai/gpt-5.4", variant: "xhigh" },
+  deep: { model: "openai/gpt-5.3-codex", variant: "medium" },
+  artistry: { model: "google/gemini-3.1-pro", variant: "high" },
+  quick: { model: "anthropic/claude-haiku-4-5" },
+  "unspecified-low": { model: "anthropic/claude-sonnet-4-6" },
+  "unspecified-high": { model: "anthropic/claude-opus-4-6", variant: "max" },
+  writing: { model: "kimi-for-coding/k2p5" },
+}
+
+export const CATEGORY_DESCRIPTIONS: Record<string, string> = {
+  "visual-engineering": "Frontend, UI/UX, design, styling, animation",
+  ultrabrain: "Use ONLY for genuinely hard, logic-heavy tasks. Give clear goals only, not step-by-step instructions.",
+  deep: "Goal-oriented autonomous problem-solving. Thorough research before action. For hairy problems requiring deep understanding.",
+  artistry: "Complex problem-solving with unconventional, creative approaches - beyond standard patterns",
+  quick: "Trivial tasks - single file changes, typo fixes, simple modifications",
+  "unspecified-low": "Tasks that don't fit other categories, low effort required",
+  "unspecified-high": "Tasks that don't fit other categories, high effort required",
+  writing: "Documentation, prose, technical writing",
+}
+```
+
+## New File: `src/tools/delegate-task/category-prompt-appends.ts`
+
+```typescript
+export const VISUAL_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on VISUAL/UI tasks.
+...
+</Category_Context>`
+// (exact content from lines 8-95 of constants.ts)
+
+export const ULTRABRAIN_CATEGORY_PROMPT_APPEND = `<Category_Context>
+...
+</Category_Context>`
+// (exact content from lines 97-117)
+
+export const ARTISTRY_CATEGORY_PROMPT_APPEND = `<Category_Context>
+...
+</Category_Context>`
+// (exact content from lines 119-134)
+
+export const QUICK_CATEGORY_PROMPT_APPEND = `<Category_Context>
+...
+</Caller_Warning>`
+// (exact content from lines 136-186)
+
+export const UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND = `<Category_Context>
+...
+</Caller_Warning>`
+// (exact content from lines 188-209)
+
+export const UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND = `<Category_Context>
+...
+</Category_Context>`
+// (exact content from lines 211-224)
+
+export const WRITING_CATEGORY_PROMPT_APPEND = `<Category_Context>
+...
+</Category_Context>`
+// (exact content from lines 226-250)
+
+export const DEEP_CATEGORY_PROMPT_APPEND = `<Category_Context>
+...
+</Category_Context>`
+// (exact content from lines 252-281)
+
+export const CATEGORY_PROMPT_APPENDS: Record<string, string> = {
+  "visual-engineering": VISUAL_CATEGORY_PROMPT_APPEND,
+  ultrabrain: ULTRABRAIN_CATEGORY_PROMPT_APPEND,
+  deep: DEEP_CATEGORY_PROMPT_APPEND,
+  artistry: ARTISTRY_CATEGORY_PROMPT_APPEND,
+  quick: QUICK_CATEGORY_PROMPT_APPEND,
+  "unspecified-low": UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND,
+  "unspecified-high": UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND,
+  writing: WRITING_CATEGORY_PROMPT_APPEND,
+}
+```
+
+## New File: `src/tools/delegate-task/plan-agent-prompt.ts`
+
+```typescript
+import type {
+  AvailableCategory,
+  AvailableSkill,
+} from "../../agents/dynamic-agent-prompt-builder"
+import { truncateDescription } from "../../shared/truncate-description"
+
+/**
+ * System prompt prepended to plan agent invocations.
+ * Instructs the plan agent to first gather context via explore/librarian agents,
+ * then summarize user requirements and clarify uncertainties before proceeding.
+ * Also MANDATES dependency graphs, parallel execution analysis, and category+skill recommendations.
+ */
+export const PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS = `<system>
+...
+</CRITICAL_REQUIREMENT_DEPENDENCY_PARALLEL_EXECUTION_CATEGORY_SKILLS>
+`
+// (exact content from lines 324-430)
+
+export const PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS = `### REQUIRED OUTPUT FORMAT
+...
+`
+// (exact content from lines 432-569)
+
+function renderPlanAgentCategoryRows(categories: AvailableCategory[]): string[] {
+  const sorted = [...categories].sort((a, b) => a.name.localeCompare(b.name))
+  return sorted.map((category) => {
+    const bestFor = category.description || category.name
+    const model = category.model || ""
+    return `| \`${category.name}\` | ${bestFor} | ${model} |`
+  })
+}
+
+function renderPlanAgentSkillRows(skills: AvailableSkill[]): string[] {
+   const sorted = [...skills].sort((a, b) => a.name.localeCompare(b.name))
+   return sorted.map((skill) => {
+     const domain = truncateDescription(skill.description).trim() || skill.name
+     return `| \`${skill.name}\` | ${domain} |`
+   })
+ }
+
+export function buildPlanAgentSkillsSection(
+  categories: AvailableCategory[] = [],
+  skills: AvailableSkill[] = []
+): string {
+  const categoryRows = renderPlanAgentCategoryRows(categories)
+  const skillRows = renderPlanAgentSkillRows(skills)
+
+  return `### AVAILABLE CATEGORIES
+
+| Category | Best For | Model |
+|----------|----------|-------|
+${categoryRows.join("\n")}
+
+### AVAILABLE SKILLS (ALWAYS EVALUATE ALL)
+
+Skills inject specialized expertise into the delegated agent.
+YOU MUST evaluate EVERY skill and justify inclusions/omissions.
+
+| Skill | Domain |
+|-------|--------|
+${skillRows.join("\n")}`
+}
+
+export function buildPlanAgentSystemPrepend(
+  categories: AvailableCategory[] = [],
+  skills: AvailableSkill[] = []
+): string {
+  return [
+    PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS,
+    buildPlanAgentSkillsSection(categories, skills),
+    PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS,
+  ].join("\n\n")
+}
+```
+
+## New File: `src/tools/delegate-task/plan-agent-names.ts`
+
+```typescript
+/**
+ * List of agent names that should be treated as plan agents (receive plan system prompt).
+ * Case-insensitive matching is used.
+ */
+export const PLAN_AGENT_NAMES = ["plan"]
+
+/**
+ * Check if the given agent name is a plan agent (receives plan system prompt).
+ */
+export function isPlanAgent(agentName: string | undefined): boolean {
+  if (!agentName) return false
+  const lowerName = agentName.toLowerCase().trim()
+  return PLAN_AGENT_NAMES.some(name => lowerName === name || lowerName.includes(name))
+}
+
+/**
+ * Plan family: plan + prometheus. Shares mutual delegation blocking and task tool permission.
+ * Does NOT share system prompt (only isPlanAgent controls that).
+ */
+export const PLAN_FAMILY_NAMES = ["plan", "prometheus"]
+
+/**
+ * Check if the given agent belongs to the plan family (blocking + task permission).
+ */
+export function isPlanFamily(category: string): boolean
+export function isPlanFamily(category: string | undefined): boolean
+export function isPlanFamily(category: string | undefined): boolean {
+  if (!category) return false
+  const lowerCategory = category.toLowerCase().trim()
+  return PLAN_FAMILY_NAMES.some(
+    (name) => lowerCategory === name || lowerCategory.includes(name)
+  )
+}
+```
+
+## Modified File: `src/tools/delegate-task/constants.ts`
+
+```typescript
+export * from "./default-categories"
+export * from "./category-prompt-appends"
+export * from "./plan-agent-prompt"
+export * from "./plan-agent-names"
+```
+
+## Unchanged: `src/tools/delegate-task/index.ts`
+
+```typescript
+export { createDelegateTask, resolveCategoryConfig, buildSystemContent, buildTaskPrompt } from "./tools"
+export type { DelegateTaskToolOptions, SyncSessionCreatedEvent, BuildSystemContentInput } from "./tools"
+export type * from "./types"
+export * from "./constants"
+```
+
+No changes needed. `export * from "./constants"` transitively re-exports everything from the 4 new files.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/with_skill/outputs/execution-plan.md

@@ -0,0 +1,104 @@
+# Execution Plan: Split delegate-task/constants.ts
+
+## Phase 0: Setup
+
+```bash
+git fetch origin dev
+git worktree add ../omo-wt/refactor-delegate-task-constants origin/dev -b refactor/split-delegate-task-constants
+cd ../omo-wt/refactor-delegate-task-constants
+```
+
+## Phase 1: Implement
+
+### Analysis
+
+`src/tools/delegate-task/constants.ts` is 654 lines with 4 distinct responsibilities:
+
+1. **Category defaults** (lines 285-316): `DEFAULT_CATEGORIES`, `CATEGORY_DESCRIPTIONS`
+2. **Category prompt appends** (lines 8-305): 8 `*_CATEGORY_PROMPT_APPEND` string constants + `CATEGORY_PROMPT_APPENDS` record
+3. **Plan agent prompts** (lines 318-620): `PLAN_AGENT_SYSTEM_PREPEND_*`, builder functions
+4. **Plan agent names** (lines 626-654): `PLAN_AGENT_NAMES`, `isPlanAgent`, `PLAN_FAMILY_NAMES`, `isPlanFamily`
+
+Note: `CATEGORY_MODEL_REQUIREMENTS` is already in `src/shared/model-requirements.ts`. No move needed.
+
+### New Files
+
+| File | Responsibility | ~LOC |
+|------|---------------|------|
+| `default-categories.ts` | `DEFAULT_CATEGORIES`, `CATEGORY_DESCRIPTIONS` | ~40 |
+| `category-prompt-appends.ts` | 8 prompt append constants + `CATEGORY_PROMPT_APPENDS` record | ~300 (exempt: prompt text) |
+| `plan-agent-prompt.ts` | Plan agent system prompt constants + builder functions | ~250 (exempt: prompt text) |
+| `plan-agent-names.ts` | `PLAN_AGENT_NAMES`, `isPlanAgent`, `PLAN_FAMILY_NAMES`, `isPlanFamily` | ~30 |
+| `constants.ts` (updated) | Re-exports from all 4 files (backward compat) | ~5 |
+
+### Commit 1: Extract category defaults and prompt appends
+
+**Files changed**: 3 new + 1 modified
+- Create `src/tools/delegate-task/default-categories.ts`
+- Create `src/tools/delegate-task/category-prompt-appends.ts`
+- Modify `src/tools/delegate-task/constants.ts` (remove extracted code, add re-exports)
+
+### Commit 2: Extract plan agent prompt and names
+
+**Files changed**: 2 new + 1 modified
+- Create `src/tools/delegate-task/plan-agent-prompt.ts`
+- Create `src/tools/delegate-task/plan-agent-names.ts`
+- Modify `src/tools/delegate-task/constants.ts` (final: re-exports only)
+
+### Local Validation
+
+```bash
+bun run typecheck
+bun test src/tools/delegate-task/
+bun run build
+```
+
+## Phase 2: PR Creation
+
+```bash
+git push -u origin refactor/split-delegate-task-constants
+gh pr create --base dev --title "refactor(delegate-task): split constants.ts into focused modules" --body-file /tmp/pr-body.md
+```
+
+## Phase 3: Verify Loop
+
+- **Gate A**: `gh pr checks --watch`
+- **Gate B**: `/review-work` (5-agent review)
+- **Gate C**: Wait for cubic-dev-ai[bot] "No issues found"
+
+## Phase 4: Merge
+
+```bash
+gh pr merge --squash --delete-branch
+git worktree remove ../omo-wt/refactor-delegate-task-constants
+```
+
+## Import Update Strategy
+
+No import updates needed. Backward compatibility preserved through:
+1. `constants.ts` re-exports everything from the 4 new files
+2. `index.ts` already does `export * from "./constants"` (unchanged)
+3. All external consumers import from `"../tools/delegate-task/constants"` or `"./constants"` -- both still work
+
+### External Import Map (Verified -- NO CHANGES NEEDED)
+
+| Consumer | Imports | Source Path |
+|----------|---------|-------------|
+| `src/agents/atlas/prompt-section-builder.ts` | `CATEGORY_DESCRIPTIONS` | `../../tools/delegate-task/constants` |
+| `src/agents/builtin-agents.ts` | `CATEGORY_DESCRIPTIONS` | `../tools/delegate-task/constants` |
+| `src/plugin/available-categories.ts` | `CATEGORY_DESCRIPTIONS` | `../tools/delegate-task/constants` |
+| `src/plugin-handlers/category-config-resolver.ts` | `DEFAULT_CATEGORIES` | `../tools/delegate-task/constants` |
+| `src/shared/merge-categories.ts` | `DEFAULT_CATEGORIES` | `../tools/delegate-task/constants` |
+| `src/shared/merge-categories.test.ts` | `DEFAULT_CATEGORIES` | `../tools/delegate-task/constants` |
+
+### Internal Import Map (Within delegate-task/ -- NO CHANGES NEEDED)
+
+| Consumer | Imports |
+|----------|---------|
+| `categories.ts` | `DEFAULT_CATEGORIES`, `CATEGORY_PROMPT_APPENDS` |
+| `tools.ts` | `CATEGORY_DESCRIPTIONS` |
+| `prompt-builder.ts` | `buildPlanAgentSystemPrepend`, `isPlanAgent` |
+| `subagent-resolver.ts` | `isPlanFamily` |
+| `sync-continuation.ts` | `isPlanFamily` |
+| `sync-prompt-sender.ts` | `isPlanFamily` |
+| `tools.test.ts` | `DEFAULT_CATEGORIES`, `CATEGORY_PROMPT_APPENDS`, `CATEGORY_DESCRIPTIONS`, `isPlanAgent`, `PLAN_AGENT_NAMES`, `isPlanFamily`, `PLAN_FAMILY_NAMES` |

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/with_skill/outputs/pr-description.md

@@ -0,0 +1,41 @@
+# PR Title
+
+```
+refactor(delegate-task): split constants.ts into focused modules
+```
+
+# PR Body
+
+## Summary
+
+- Split the 654-line `src/tools/delegate-task/constants.ts` into 4 single-responsibility modules: `default-categories.ts`, `category-prompt-appends.ts`, `plan-agent-prompt.ts`, `plan-agent-names.ts`
+- `constants.ts` becomes a pure re-export barrel, preserving all existing import paths (`from "./constants"` and `from "./delegate-task"`)
+- Zero import changes across the codebase (6 external + 7 internal consumers verified)
+
+## Motivation
+
+`constants.ts` at 654 lines violates the project's 200 LOC soft limit (`modular-code-enforcement.md` rule) and bundles 4 unrelated responsibilities: category model configs, category prompt text, plan agent prompts, and plan agent name utilities.
+
+## Changes
+
+| New File | Responsibility | LOC |
+|----------|---------------|-----|
+| `default-categories.ts` | `DEFAULT_CATEGORIES`, `CATEGORY_DESCRIPTIONS` | ~25 |
+| `category-prompt-appends.ts` | 8 `*_PROMPT_APPEND` constants + `CATEGORY_PROMPT_APPENDS` record | ~300 (prompt-exempt) |
+| `plan-agent-prompt.ts` | Plan system prompt constants + `buildPlanAgentSystemPrepend()` | ~250 (prompt-exempt) |
+| `plan-agent-names.ts` | `PLAN_AGENT_NAMES`, `isPlanAgent`, `PLAN_FAMILY_NAMES`, `isPlanFamily` | ~30 |
+| `constants.ts` (updated) | 4-line re-export barrel | 4 |
+
+## Backward Compatibility
+
+All 13 consumers continue importing from `"./constants"` or `"../tools/delegate-task/constants"` with zero changes. The re-export chain: new modules -> `constants.ts` -> `index.ts` -> external consumers.
+
+## Note on CATEGORY_MODEL_REQUIREMENTS
+
+`CATEGORY_MODEL_REQUIREMENTS` already lives in `src/shared/model-requirements.ts`. No move needed. The AGENTS.md reference to it being in `constants.ts` is outdated.
+
+## Testing
+
+- `bun run typecheck` passes
+- `bun test src/tools/delegate-task/` passes (all existing tests untouched)
+- `bun run build` succeeds

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/with_skill/outputs/verification-strategy.md

@@ -0,0 +1,84 @@
+# Verification Strategy
+
+## Gate A: CI (Blocking)
+
+```bash
+gh pr checks --watch
+```
+
+**Expected CI jobs** (from `ci.yml`):
+1. **Tests (split)**: mock-heavy isolated + batch `bun test`
+2. **Typecheck**: `bun run typecheck` (tsc --noEmit)
+3. **Build**: `bun run build`
+4. **Schema auto-commit**: If schema changes detected
+
+**Likely failure points**: None. This is a pure refactor with re-exports. No runtime behavior changes.
+
+**If CI fails**:
+- Typecheck error: Missing re-export or import cycle. Fix in the new modules, amend commit.
+- Test error: `tools.test.ts` imports all symbols from `"./constants"`. Re-export barrel must be complete.
+
+## Gate B: review-work (5-Agent Review)
+
+Invoke after CI passes:
+
+```
+/review-work
+```
+
+**5 parallel agents**:
+1. **Oracle (goal/constraint)**: Verify backward compat claim. Check all 13 import paths resolve.
+2. **Oracle (code quality)**: Verify single-responsibility per file, LOC limits, no catch-all violations.
+3. **Oracle (security)**: No security implications in this refactor.
+4. **QA (hands-on execution)**: Run `bun test src/tools/delegate-task/` and verify all pass.
+5. **Context miner**: Check no related open issues/PRs conflict.
+
+**Expected verdict**: Pass. Pure structural refactor with no behavioral changes.
+
+## Gate C: Cubic (External Bot)
+
+Wait for `cubic-dev-ai[bot]` to post "No issues found" on the PR.
+
+**If Cubic flags issues**: Likely false positives on "large number of new files". Address in PR comments if needed.
+
+## Pre-Gate Local Validation (Before Push)
+
+```bash
+# In worktree
+bun run typecheck
+bun test src/tools/delegate-task/
+bun run build
+
+# Verify re-exports are complete
+bun -e "import * as c from './src/tools/delegate-task/constants'; console.log(Object.keys(c).sort().join('\n'))"
+```
+
+Expected exports from constants.ts (13 total):
+- `ARTISTRY_CATEGORY_PROMPT_APPEND`
+- `CATEGORY_DESCRIPTIONS`
+- `CATEGORY_PROMPT_APPENDS`
+- `DEFAULT_CATEGORIES`
+- `DEEP_CATEGORY_PROMPT_APPEND`
+- `PLAN_AGENT_NAMES`
+- `PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS`
+- `PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS`
+- `PLAN_FAMILY_NAMES`
+- `QUICK_CATEGORY_PROMPT_APPEND`
+- `ULTRABRAIN_CATEGORY_PROMPT_APPEND`
+- `UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND`
+- `UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND`
+- `VISUAL_CATEGORY_PROMPT_APPEND`
+- `WRITING_CATEGORY_PROMPT_APPEND`
+- `buildPlanAgentSkillsSection`
+- `buildPlanAgentSystemPrepend`
+- `isPlanAgent`
+- `isPlanFamily`
+
+## Merge Strategy
+
+```bash
+gh pr merge --squash --delete-branch
+git worktree remove ../omo-wt/refactor-delegate-task-constants
+```
+
+Squash merge collapses the 2 atomic commits into 1 clean commit on dev.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/with_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 181000, "total_duration_seconds": 181}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/without_skill/grading.json

@@ -0,0 +1,10 @@
+{
+  "run_id": "eval-3-without_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": false, "evidence": "git checkout -b only, no worktree"},
+    {"text": "Uses 2+ commits for the multi-file refactor", "passed": false, "evidence": "Single atomic commit: 'refactor: split delegate-task constants and category model requirements'"},
+    {"text": "Maintains backward compatibility via barrel re-export", "passed": true, "evidence": "Re-exports from new files, zero consumer changes"},
+    {"text": "Verification loop includes all 3 gates", "passed": false, "evidence": "Only mentions typecheck/test/build. No review-work or Cubic."},
+    {"text": "References actual src/tools/delegate-task/constants.ts", "passed": true, "evidence": "654 lines, detailed responsibility breakdown, full import maps"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/without_skill/outputs/code-changes.md

@@ -0,0 +1,342 @@
+# Code Changes
+
+## 1. NEW: `src/tools/delegate-task/default-categories.ts`
+
+```typescript
+import type { CategoryConfig } from "../../config/schema"
+
+export const DEFAULT_CATEGORIES: Record<string, CategoryConfig> = {
+  "visual-engineering": { model: "google/gemini-3.1-pro", variant: "high" },
+  ultrabrain: { model: "openai/gpt-5.4", variant: "xhigh" },
+  deep: { model: "openai/gpt-5.3-codex", variant: "medium" },
+  artistry: { model: "google/gemini-3.1-pro", variant: "high" },
+  quick: { model: "anthropic/claude-haiku-4-5" },
+  "unspecified-low": { model: "anthropic/claude-sonnet-4-6" },
+  "unspecified-high": { model: "anthropic/claude-opus-4-6", variant: "max" },
+  writing: { model: "kimi-for-coding/k2p5" },
+}
+```
+
+## 2. NEW: `src/tools/delegate-task/category-descriptions.ts`
+
+```typescript
+export const CATEGORY_DESCRIPTIONS: Record<string, string> = {
+  "visual-engineering": "Frontend, UI/UX, design, styling, animation",
+  ultrabrain: "Use ONLY for genuinely hard, logic-heavy tasks. Give clear goals only, not step-by-step instructions.",
+  deep: "Goal-oriented autonomous problem-solving. Thorough research before action. For hairy problems requiring deep understanding.",
+  artistry: "Complex problem-solving with unconventional, creative approaches - beyond standard patterns",
+  quick: "Trivial tasks - single file changes, typo fixes, simple modifications",
+  "unspecified-low": "Tasks that don't fit other categories, low effort required",
+  "unspecified-high": "Tasks that don't fit other categories, high effort required",
+  writing: "Documentation, prose, technical writing",
+}
+```
+
+## 3. NEW: `src/tools/delegate-task/category-prompt-appends.ts`
+
+```typescript
+export const VISUAL_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on VISUAL/UI tasks.
+...
+</Category_Context>`
+
+export const ULTRABRAIN_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on DEEP LOGICAL REASONING / COMPLEX ARCHITECTURE tasks.
+...
+</Category_Context>`
+
+export const ARTISTRY_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on HIGHLY CREATIVE / ARTISTIC tasks.
+...
+</Category_Context>`
+
+export const QUICK_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on SMALL / QUICK tasks.
+...
+</Caller_Warning>`
+
+export const UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on tasks that don't fit specific categories but require moderate effort.
+...
+</Caller_Warning>`
+
+export const UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on tasks that don't fit specific categories but require substantial effort.
+...
+</Category_Context>`
+
+export const WRITING_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on WRITING / PROSE tasks.
+...
+</Category_Context>`
+
+export const DEEP_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on GOAL-ORIENTED AUTONOMOUS tasks.
+...
+</Category_Context>`
+
+export const CATEGORY_PROMPT_APPENDS: Record<string, string> = {
+  "visual-engineering": VISUAL_CATEGORY_PROMPT_APPEND,
+  ultrabrain: ULTRABRAIN_CATEGORY_PROMPT_APPEND,
+  deep: DEEP_CATEGORY_PROMPT_APPEND,
+  artistry: ARTISTRY_CATEGORY_PROMPT_APPEND,
+  quick: QUICK_CATEGORY_PROMPT_APPEND,
+  "unspecified-low": UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND,
+  "unspecified-high": UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND,
+  writing: WRITING_CATEGORY_PROMPT_APPEND,
+}
+```
+
+> Note: Each `*_CATEGORY_PROMPT_APPEND` contains the full template string from the original. Abbreviated with `...` here for readability. The actual code would contain the complete unmodified prompt text.
+
+## 4. NEW: `src/tools/delegate-task/plan-agent-prompt.ts`
+
+```typescript
+import type {
+  AvailableCategory,
+  AvailableSkill,
+} from "../../agents/dynamic-agent-prompt-builder"
+import { truncateDescription } from "../../shared/truncate-description"
+
+export const PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS = `<system>
+BEFORE you begin planning, you MUST first understand the user's request deeply.
+...
+</CRITICAL_REQUIREMENT_DEPENDENCY_PARALLEL_EXECUTION_CATEGORY_SKILLS>
+
+<FINAL_OUTPUT_FOR_CALLER>
+...
+</FINAL_OUTPUT_FOR_CALLER>
+
+`
+
+export const PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS = `### REQUIRED OUTPUT FORMAT
+...
+`
+
+function renderPlanAgentCategoryRows(categories: AvailableCategory[]): string[] {
+  const sorted = [...categories].sort((a, b) => a.name.localeCompare(b.name))
+  return sorted.map((category) => {
+    const bestFor = category.description || category.name
+    const model = category.model || ""
+    return `| \`${category.name}\` | ${bestFor} | ${model} |`
+  })
+}
+
+function renderPlanAgentSkillRows(skills: AvailableSkill[]): string[] {
+   const sorted = [...skills].sort((a, b) => a.name.localeCompare(b.name))
+   return sorted.map((skill) => {
+     const domain = truncateDescription(skill.description).trim() || skill.name
+     return `| \`${skill.name}\` | ${domain} |`
+   })
+ }
+
+export function buildPlanAgentSkillsSection(
+  categories: AvailableCategory[] = [],
+  skills: AvailableSkill[] = []
+): string {
+  const categoryRows = renderPlanAgentCategoryRows(categories)
+  const skillRows = renderPlanAgentSkillRows(skills)
+
+  return `### AVAILABLE CATEGORIES
+
+| Category | Best For | Model |
+|----------|----------|-------|
+${categoryRows.join("\n")}
+
+### AVAILABLE SKILLS (ALWAYS EVALUATE ALL)
+
+Skills inject specialized expertise into the delegated agent.
+YOU MUST evaluate EVERY skill and justify inclusions/omissions.
+
+| Skill | Domain |
+|-------|--------|
+${skillRows.join("\n")}`
+}
+
+export function buildPlanAgentSystemPrepend(
+  categories: AvailableCategory[] = [],
+  skills: AvailableSkill[] = []
+): string {
+  return [
+    PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS,
+    buildPlanAgentSkillsSection(categories, skills),
+    PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS,
+  ].join("\n\n")
+}
+```
+
+> Note: Template strings abbreviated with `...`. Full unmodified content in the actual file.
+
+## 5. NEW: `src/tools/delegate-task/plan-agent-identity.ts`
+
+```typescript
+/**
+ * List of agent names that should be treated as plan agents (receive plan system prompt).
+ * Case-insensitive matching is used.
+ */
+export const PLAN_AGENT_NAMES = ["plan"]
+
+/**
+ * Check if the given agent name is a plan agent (receives plan system prompt).
+ */
+export function isPlanAgent(agentName: string | undefined): boolean {
+  if (!agentName) return false
+  const lowerName = agentName.toLowerCase().trim()
+  return PLAN_AGENT_NAMES.some(name => lowerName === name || lowerName.includes(name))
+}
+
+/**
+ * Plan family: plan + prometheus. Shares mutual delegation blocking and task tool permission.
+ * Does NOT share system prompt (only isPlanAgent controls that).
+ */
+export const PLAN_FAMILY_NAMES = ["plan", "prometheus"]
+
+/**
+ * Check if the given agent belongs to the plan family (blocking + task permission).
+ */
+export function isPlanFamily(category: string): boolean
+export function isPlanFamily(category: string | undefined): boolean
+export function isPlanFamily(category: string | undefined): boolean {
+  if (!category) return false
+  const lowerCategory = category.toLowerCase().trim()
+  return PLAN_FAMILY_NAMES.some(
+    (name) => lowerCategory === name || lowerCategory.includes(name)
+  )
+}
+```
+
+## 6. MODIFIED: `src/tools/delegate-task/constants.ts` (barrel re-export)
+
+```typescript
+export { DEFAULT_CATEGORIES } from "./default-categories"
+export { CATEGORY_DESCRIPTIONS } from "./category-descriptions"
+export {
+  VISUAL_CATEGORY_PROMPT_APPEND,
+  ULTRABRAIN_CATEGORY_PROMPT_APPEND,
+  ARTISTRY_CATEGORY_PROMPT_APPEND,
+  QUICK_CATEGORY_PROMPT_APPEND,
+  UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND,
+  UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND,
+  WRITING_CATEGORY_PROMPT_APPEND,
+  DEEP_CATEGORY_PROMPT_APPEND,
+  CATEGORY_PROMPT_APPENDS,
+} from "./category-prompt-appends"
+export {
+  PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS,
+  PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS,
+  buildPlanAgentSkillsSection,
+  buildPlanAgentSystemPrepend,
+} from "./plan-agent-prompt"
+export {
+  PLAN_AGENT_NAMES,
+  isPlanAgent,
+  PLAN_FAMILY_NAMES,
+  isPlanFamily,
+} from "./plan-agent-identity"
+```
+
+## 7. NEW: `src/shared/category-model-requirements.ts`
+
+```typescript
+import type { ModelRequirement } from "./model-requirements"
+
+export const CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
+  "visual-engineering": {
+    fallbackChain: [
+      {
+        providers: ["google", "github-copilot", "opencode"],
+        model: "gemini-3.1-pro",
+        variant: "high",
+      },
+      { providers: ["zai-coding-plan", "opencode"], model: "glm-5" },
+      {
+        providers: ["anthropic", "github-copilot", "opencode"],
+        model: "claude-opus-4-6",
+        variant: "max",
+      },
+      { providers: ["opencode-go"], model: "glm-5" },
+      { providers: ["kimi-for-coding"], model: "k2p5" },
+    ],
+  },
+  ultrabrain: {
+    fallbackChain: [
+      // ... full content from original
+    ],
+  },
+  deep: {
+    fallbackChain: [
+      // ... full content from original
+    ],
+    requiresModel: "gpt-5.3-codex",
+  },
+  artistry: {
+    fallbackChain: [
+      // ... full content from original
+    ],
+    requiresModel: "gemini-3.1-pro",
+  },
+  quick: {
+    fallbackChain: [
+      // ... full content from original
+    ],
+  },
+  "unspecified-low": {
+    fallbackChain: [
+      // ... full content from original
+    ],
+  },
+  "unspecified-high": {
+    fallbackChain: [
+      // ... full content from original
+    ],
+  },
+  writing: {
+    fallbackChain: [
+      // ... full content from original
+    ],
+  },
+}
+```
+
+> Note: Each category's `fallbackChain` contains the exact same entries as the original `model-requirements.ts`. Abbreviated here.
+
+## 8. MODIFIED: `src/shared/model-requirements.ts`
+
+**Remove** `CATEGORY_MODEL_REQUIREMENTS` from the file body. **Add** re-export at the end:
+
+```typescript
+export type FallbackEntry = {
+  providers: string[];
+  model: string;
+  variant?: string;
+};
+
+export type ModelRequirement = {
+  fallbackChain: FallbackEntry[];
+  variant?: string;
+  requiresModel?: string;
+  requiresAnyModel?: boolean;
+  requiresProvider?: string[];
+};
+
+export const AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
+  // ... unchanged, full agent entries stay here
+};
+
+export { CATEGORY_MODEL_REQUIREMENTS } from "./category-model-requirements"
+```
+
+## Summary of Changes
+
+| File | Lines Before | Lines After | Action |
+|------|-------------|-------------|--------|
+| `constants.ts` | 654 | ~25 | Rewrite as barrel re-export |
+| `default-categories.ts` | - | ~15 | **NEW** |
+| `category-descriptions.ts` | - | ~12 | **NEW** |
+| `category-prompt-appends.ts` | - | ~280 | **NEW** (mostly exempt prompt text) |
+| `plan-agent-prompt.ts` | - | ~270 | **NEW** (mostly exempt prompt text) |
+| `plan-agent-identity.ts` | - | ~35 | **NEW** |
+| `model-requirements.ts` | 311 | ~165 | Remove CATEGORY_MODEL_REQUIREMENTS |
+| `category-model-requirements.ts` | - | ~150 | **NEW** |
+
+**Zero consumer files modified.** Backward compatibility maintained through barrel re-exports.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/without_skill/outputs/execution-plan.md

@@ -0,0 +1,131 @@
+# Execution Plan: Refactor constants.ts
+
+## Context
+
+`src/tools/delegate-task/constants.ts` is **654 lines** with 6 distinct responsibilities. Violates the 200 LOC modular-code-enforcement rule. `CATEGORY_MODEL_REQUIREMENTS` is actually in `src/shared/model-requirements.ts` (311 lines, also violating 200 LOC), not in `constants.ts`.
+
+## Pre-Flight Analysis
+
+### Current `constants.ts` responsibilities:
+1. **Category prompt appends** (8 template strings, ~274 LOC prompt text)
+2. **DEFAULT_CATEGORIES** (Record<string, CategoryConfig>, ~10 LOC)
+3. **CATEGORY_PROMPT_APPENDS** (map of category->prompt, ~10 LOC)
+4. **CATEGORY_DESCRIPTIONS** (map of category->description, ~10 LOC)
+5. **Plan agent prompts** (2 template strings + 4 builder functions, ~250 LOC prompt text)
+6. **Plan agent identity utils** (`isPlanAgent`, `isPlanFamily`, ~30 LOC)
+
+### Current `model-requirements.ts` responsibilities:
+1. Types (`FallbackEntry`, `ModelRequirement`)
+2. `AGENT_MODEL_REQUIREMENTS` (~146 LOC)
+3. `CATEGORY_MODEL_REQUIREMENTS` (~148 LOC)
+
+### Import dependency map for `constants.ts`:
+
+**Internal consumers (within delegate-task/):**
+| File | Imports |
+|------|---------|
+| `categories.ts` | `DEFAULT_CATEGORIES`, `CATEGORY_PROMPT_APPENDS` |
+| `tools.ts` | `CATEGORY_DESCRIPTIONS` |
+| `tools.test.ts` | `DEFAULT_CATEGORIES`, `CATEGORY_PROMPT_APPENDS`, `CATEGORY_DESCRIPTIONS`, `isPlanAgent`, `PLAN_AGENT_NAMES`, `isPlanFamily`, `PLAN_FAMILY_NAMES` |
+| `prompt-builder.ts` | `buildPlanAgentSystemPrepend`, `isPlanAgent` |
+| `subagent-resolver.ts` | `isPlanFamily` |
+| `sync-continuation.ts` | `isPlanFamily` |
+| `sync-prompt-sender.ts` | `isPlanFamily` |
+| `index.ts` | `export * from "./constants"` (barrel) |
+
+**External consumers (import from `"../../tools/delegate-task/constants"`):**
+| File | Imports |
+|------|---------|
+| `agents/atlas/prompt-section-builder.ts` | `CATEGORY_DESCRIPTIONS` |
+| `agents/builtin-agents.ts` | `CATEGORY_DESCRIPTIONS` |
+| `plugin/available-categories.ts` | `CATEGORY_DESCRIPTIONS` |
+| `plugin-handlers/category-config-resolver.ts` | `DEFAULT_CATEGORIES` |
+| `shared/merge-categories.ts` | `DEFAULT_CATEGORIES` |
+| `shared/merge-categories.test.ts` | `DEFAULT_CATEGORIES` |
+
+**External consumers of `CATEGORY_MODEL_REQUIREMENTS`:**
+| File | Import path |
+|------|-------------|
+| `tools/delegate-task/categories.ts` | `../../shared/model-requirements` |
+
+## Step-by-Step Execution
+
+### Step 1: Create branch
+```bash
+git checkout -b refactor/split-category-constants dev
+```
+
+### Step 2: Split `constants.ts` into 5 focused files
+
+#### 2a. Create `default-categories.ts`
+- Move `DEFAULT_CATEGORIES` record
+- Import `CategoryConfig` type from config schema
+- ~15 LOC
+
+#### 2b. Create `category-descriptions.ts`
+- Move `CATEGORY_DESCRIPTIONS` record
+- No dependencies
+- ~12 LOC
+
+#### 2c. Create `category-prompt-appends.ts`
+- Move all 8 `*_CATEGORY_PROMPT_APPEND` template string constants
+- Move `CATEGORY_PROMPT_APPENDS` mapping record
+- No dependencies (all self-contained template strings)
+- ~280 LOC (mostly prompt text, exempt from 200 LOC per modular-code-enforcement)
+
+#### 2d. Create `plan-agent-prompt.ts`
+- Move `PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS`
+- Move `PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS`
+- Move `renderPlanAgentCategoryRows()`, `renderPlanAgentSkillRows()`
+- Move `buildPlanAgentSkillsSection()`, `buildPlanAgentSystemPrepend()`
+- Imports: `AvailableCategory`, `AvailableSkill` from agents, `truncateDescription` from shared
+- ~270 LOC (mostly prompt text, exempt)
+
+#### 2e. Create `plan-agent-identity.ts`
+- Move `PLAN_AGENT_NAMES`, `isPlanAgent()`
+- Move `PLAN_FAMILY_NAMES`, `isPlanFamily()`
+- No dependencies
+- ~35 LOC
+
+### Step 3: Convert `constants.ts` to barrel re-export file
+Replace entire contents with re-exports from the 5 new files. This maintains 100% backward compatibility for all existing importers.
+
+### Step 4: Split `model-requirements.ts`
+
+#### 4a. Create `src/shared/category-model-requirements.ts`
+- Move `CATEGORY_MODEL_REQUIREMENTS` record
+- Import `ModelRequirement` type from `./model-requirements`
+- ~150 LOC
+
+#### 4b. Update `model-requirements.ts`
+- Remove `CATEGORY_MODEL_REQUIREMENTS`
+- Add re-export: `export { CATEGORY_MODEL_REQUIREMENTS } from "./category-model-requirements"`
+- Keep types (`FallbackEntry`, `ModelRequirement`) and `AGENT_MODEL_REQUIREMENTS`
+- ~165 LOC (now under 200)
+
+### Step 5: Verify no import breakage
+- Run `bun run typecheck` to confirm all imports resolve
+- Run `bun test` to confirm no behavioral regressions
+- Run `bun run build` to confirm build succeeds
+
+### Step 6: Verify LSP diagnostics clean
+- Check `lsp_diagnostics` on all new and modified files
+
+### Step 7: Commit and create PR
+- Single atomic commit: `refactor: split delegate-task constants and category model requirements into focused modules`
+- Create PR with description
+
+## Files Modified
+
+| File | Action |
+|------|--------|
+| `src/tools/delegate-task/constants.ts` | Rewrite as barrel re-export |
+| `src/tools/delegate-task/default-categories.ts` | **NEW** |
+| `src/tools/delegate-task/category-descriptions.ts` | **NEW** |
+| `src/tools/delegate-task/category-prompt-appends.ts` | **NEW** |
+| `src/tools/delegate-task/plan-agent-prompt.ts` | **NEW** |
+| `src/tools/delegate-task/plan-agent-identity.ts` | **NEW** |
+| `src/shared/model-requirements.ts` | Remove CATEGORY_MODEL_REQUIREMENTS, add re-export |
+| `src/shared/category-model-requirements.ts` | **NEW** |
+
+**Zero changes to any consumer files.** All existing imports work via barrel re-exports.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/without_skill/outputs/pr-description.md

@@ -0,0 +1,39 @@
+## Summary
+
+- Split `src/tools/delegate-task/constants.ts` (654 LOC, 6 responsibilities) into 5 focused modules: `default-categories.ts`, `category-descriptions.ts`, `category-prompt-appends.ts`, `plan-agent-prompt.ts`, `plan-agent-identity.ts`
+- Extract `CATEGORY_MODEL_REQUIREMENTS` from `src/shared/model-requirements.ts` (311 LOC) into `category-model-requirements.ts`, bringing both files under the 200 LOC limit
+- Convert original files to barrel re-exports for 100% backward compatibility (zero consumer changes)
+
+## Motivation
+
+Both files violate the project's 200 LOC modular-code-enforcement rule. `constants.ts` mixed 6 unrelated responsibilities (category configs, prompt templates, plan agent builders, identity utils). `model-requirements.ts` mixed agent and category model requirements.
+
+## Changes
+
+### `src/tools/delegate-task/`
+| New File | Responsibility |
+|----------|---------------|
+| `default-categories.ts` | `DEFAULT_CATEGORIES` record |
+| `category-descriptions.ts` | `CATEGORY_DESCRIPTIONS` record |
+| `category-prompt-appends.ts` | 8 prompt template constants + `CATEGORY_PROMPT_APPENDS` map |
+| `plan-agent-prompt.ts` | Plan agent system prompts + builder functions |
+| `plan-agent-identity.ts` | `isPlanAgent`, `isPlanFamily` + name lists |
+
+`constants.ts` is now a barrel re-export file (~25 LOC).
+
+### `src/shared/`
+| New File | Responsibility |
+|----------|---------------|
+| `category-model-requirements.ts` | `CATEGORY_MODEL_REQUIREMENTS` record |
+
+`model-requirements.ts` retains types + `AGENT_MODEL_REQUIREMENTS` and re-exports `CATEGORY_MODEL_REQUIREMENTS`.
+
+## Backward Compatibility
+
+All existing import paths (`from "./constants"`, `from "../../tools/delegate-task/constants"`, `from "../../shared/model-requirements"`) continue to work unchanged. Zero consumer files modified.
+
+## Testing
+
+- `bun run typecheck` passes
+- `bun test` passes (existing `tools.test.ts` validates all re-exported symbols)
+- `bun run build` succeeds

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/without_skill/outputs/verification-strategy.md

@@ -0,0 +1,128 @@
+# Verification Strategy
+
+## 1. Type Safety
+
+### 1a. LSP diagnostics on all new files
+```
+lsp_diagnostics("src/tools/delegate-task/default-categories.ts")
+lsp_diagnostics("src/tools/delegate-task/category-descriptions.ts")
+lsp_diagnostics("src/tools/delegate-task/category-prompt-appends.ts")
+lsp_diagnostics("src/tools/delegate-task/plan-agent-prompt.ts")
+lsp_diagnostics("src/tools/delegate-task/plan-agent-identity.ts")
+lsp_diagnostics("src/shared/category-model-requirements.ts")
+```
+
+### 1b. LSP diagnostics on modified files
+```
+lsp_diagnostics("src/tools/delegate-task/constants.ts")
+lsp_diagnostics("src/shared/model-requirements.ts")
+```
+
+### 1c. Full typecheck
+```bash
+bun run typecheck
+```
+Expected: 0 errors. This confirms all 14 consumer files (8 internal + 6 external) resolve their imports correctly through the barrel re-exports.
+
+## 2. Behavioral Regression
+
+### 2a. Existing test suite
+```bash
+bun test src/tools/delegate-task/tools.test.ts
+```
+This test file imports `DEFAULT_CATEGORIES`, `CATEGORY_PROMPT_APPENDS`, `CATEGORY_DESCRIPTIONS`, `isPlanAgent`, `PLAN_AGENT_NAMES`, `isPlanFamily`, `PLAN_FAMILY_NAMES` from `./constants`. If the barrel re-export is correct, all these tests pass unchanged.
+
+### 2b. Category resolver tests
+```bash
+bun test src/tools/delegate-task/category-resolver.test.ts
+```
+This exercises `resolveCategoryConfig()` which imports `DEFAULT_CATEGORIES` and `CATEGORY_PROMPT_APPENDS` from `./constants` and `CATEGORY_MODEL_REQUIREMENTS` from `../../shared/model-requirements`.
+
+### 2c. Model selection tests
+```bash
+bun test src/tools/delegate-task/model-selection.test.ts
+```
+
+### 2d. Merge categories tests
+```bash
+bun test src/shared/merge-categories.test.ts
+```
+Imports `DEFAULT_CATEGORIES` from `../tools/delegate-task/constants` (external path).
+
+### 2e. Full test suite
+```bash
+bun test
+```
+
+## 3. Build Verification
+
+```bash
+bun run build
+```
+Confirms ESM bundle + declarations emit correctly with the new file structure.
+
+## 4. Export Completeness Verification
+
+### 4a. Verify `constants.ts` re-exports match original exports
+Cross-check that every symbol previously exported from `constants.ts` is still exported. The original file exported these symbols:
+- `VISUAL_CATEGORY_PROMPT_APPEND`
+- `ULTRABRAIN_CATEGORY_PROMPT_APPEND`
+- `ARTISTRY_CATEGORY_PROMPT_APPEND`
+- `QUICK_CATEGORY_PROMPT_APPEND`
+- `UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND`
+- `UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND`
+- `WRITING_CATEGORY_PROMPT_APPEND`
+- `DEEP_CATEGORY_PROMPT_APPEND`
+- `DEFAULT_CATEGORIES`
+- `CATEGORY_PROMPT_APPENDS`
+- `CATEGORY_DESCRIPTIONS`
+- `PLAN_AGENT_SYSTEM_PREPEND_STATIC_BEFORE_SKILLS`
+- `PLAN_AGENT_SYSTEM_PREPEND_STATIC_AFTER_SKILLS`
+- `buildPlanAgentSkillsSection`
+- `buildPlanAgentSystemPrepend`
+- `PLAN_AGENT_NAMES`
+- `isPlanAgent`
+- `PLAN_FAMILY_NAMES`
+- `isPlanFamily`
+
+All 19 must be re-exported from the barrel.
+
+### 4b. Verify `model-requirements.ts` re-exports match original exports
+Original exports: `FallbackEntry`, `ModelRequirement`, `AGENT_MODEL_REQUIREMENTS`, `CATEGORY_MODEL_REQUIREMENTS`. All 4 must still be available.
+
+## 5. LOC Compliance Check
+
+Verify each new file is under 200 LOC (excluding prompt template text per modular-code-enforcement rule):
+
+| File | Expected Total LOC | Non-prompt LOC | Compliant? |
+|------|-------------------|----------------|------------|
+| `default-categories.ts` | ~15 | ~15 | Yes |
+| `category-descriptions.ts` | ~12 | ~12 | Yes |
+| `category-prompt-appends.ts` | ~280 | ~15 | Yes (prompt exempt) |
+| `plan-agent-prompt.ts` | ~270 | ~40 | Yes (prompt exempt) |
+| `plan-agent-identity.ts` | ~35 | ~35 | Yes |
+| `category-model-requirements.ts` | ~150 | ~150 | Yes |
+| `model-requirements.ts` (after) | ~165 | ~165 | Yes |
+| `constants.ts` (after) | ~25 | ~25 | Yes |
+
+## 6. Consumer Impact Matrix
+
+Verify zero consumer files need changes:
+
+| Consumer File | Import Path | Should Still Work? |
+|--------------|-------------|-------------------|
+| `delegate-task/categories.ts` | `./constants` | Yes (barrel) |
+| `delegate-task/tools.ts` | `./constants` | Yes (barrel) |
+| `delegate-task/tools.test.ts` | `./constants` | Yes (barrel) |
+| `delegate-task/prompt-builder.ts` | `./constants` | Yes (barrel) |
+| `delegate-task/subagent-resolver.ts` | `./constants` | Yes (barrel) |
+| `delegate-task/sync-continuation.ts` | `./constants` | Yes (barrel) |
+| `delegate-task/sync-prompt-sender.ts` | `./constants` | Yes (barrel) |
+| `delegate-task/index.ts` | `./constants` | Yes (barrel) |
+| `agents/atlas/prompt-section-builder.ts` | `../../tools/delegate-task/constants` | Yes (barrel) |
+| `agents/builtin-agents.ts` | `../tools/delegate-task/constants` | Yes (barrel) |
+| `plugin/available-categories.ts` | `../tools/delegate-task/constants` | Yes (barrel) |
+| `plugin-handlers/category-config-resolver.ts` | `../tools/delegate-task/constants` | Yes (barrel) |
+| `shared/merge-categories.ts` | `../tools/delegate-task/constants` | Yes (barrel) |
+| `shared/merge-categories.test.ts` | `../tools/delegate-task/constants` | Yes (barrel) |
+| `delegate-task/categories.ts` | `../../shared/model-requirements` | Yes (re-export) |

.opencode/skills/work-with-pr-workspace/iteration-1/eval-3/without_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 229000, "total_duration_seconds": 229}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/eval_metadata.json

@@ -0,0 +1,32 @@
+{
+  "eval_id": 4,
+  "eval_name": "new-mcp-arxiv-casual",
+  "prompt": "implement issue #100 - we need to add a new built-in MCP for arxiv paper search. just the basic search endpoint, nothing fancy. pr it",
+  "assertions": [
+    {
+      "id": "worktree-isolation",
+      "text": "Plan uses git worktree in a sibling directory",
+      "type": "manual"
+    },
+    {
+      "id": "follows-mcp-pattern",
+      "text": "New MCP follows existing pattern from src/mcp/ (websearch, context7, grep_app)",
+      "type": "manual"
+    },
+    {
+      "id": "three-gates",
+      "text": "Verification loop includes all 3 gates",
+      "type": "manual"
+    },
+    {
+      "id": "pr-targets-dev",
+      "text": "PR targets dev branch",
+      "type": "manual"
+    },
+    {
+      "id": "local-validation",
+      "text": "Runs local checks before pushing",
+      "type": "manual"
+    }
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/with_skill/grading.json

@@ -0,0 +1,10 @@
+{
+  "run_id": "eval-4-with_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": true, "evidence": "../omo-wt/feat/arxiv-mcp"},
+    {"text": "New MCP follows existing pattern from src/mcp/", "passed": true, "evidence": "Follows context7.ts and grep-app.ts static export pattern"},
+    {"text": "Verification loop includes all 3 gates", "passed": true, "evidence": "Gate A (CI), Gate B (review-work 5 agents), Gate C (Cubic)"},
+    {"text": "PR targets dev branch", "passed": true, "evidence": "--base dev"},
+    {"text": "Runs local checks before pushing", "passed": true, "evidence": "bun run typecheck, bun test src/mcp/, bun run build"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/with_skill/outputs/code-changes.md

@@ -0,0 +1,143 @@
+# Code Changes: Issue #100 - Built-in arXiv MCP
+
+## 1. NEW FILE: `src/mcp/arxiv.ts`
+
+```typescript
+export const arxiv = {
+  type: "remote" as const,
+  url: "https://mcp.arxiv.org",
+  enabled: true,
+  oauth: false as const,
+}
+```
+
+Pattern: identical to `grep-app.ts` (static export, no auth, no config factory needed).
+
+## 2. MODIFY: `src/mcp/types.ts`
+
+```typescript
+import { z } from "zod"
+
+export const McpNameSchema = z.enum(["websearch", "context7", "grep_app", "arxiv"])
+
+export type McpName = z.infer<typeof McpNameSchema>
+
+export const AnyMcpNameSchema = z.string().min(1)
+
+export type AnyMcpName = z.infer<typeof AnyMcpNameSchema>
+```
+
+Change: add `"arxiv"` to `McpNameSchema` enum.
+
+## 3. MODIFY: `src/mcp/index.ts`
+
+```typescript
+import { createWebsearchConfig } from "./websearch"
+import { context7 } from "./context7"
+import { grep_app } from "./grep-app"
+import { arxiv } from "./arxiv"
+import type { OhMyOpenCodeConfig } from "../config/schema"
+
+export { McpNameSchema, type McpName } from "./types"
+
+type RemoteMcpConfig = {
+  type: "remote"
+  url: string
+  enabled: boolean
+  headers?: Record<string, string>
+  oauth?: false
+}
+
+export function createBuiltinMcps(disabledMcps: string[] = [], config?: OhMyOpenCodeConfig) {
+  const mcps: Record<string, RemoteMcpConfig> = {}
+
+  if (!disabledMcps.includes("websearch")) {
+    mcps.websearch = createWebsearchConfig(config?.websearch)
+  }
+
+  if (!disabledMcps.includes("context7")) {
+    mcps.context7 = context7
+  }
+
+  if (!disabledMcps.includes("grep_app")) {
+    mcps.grep_app = grep_app
+  }
+
+  if (!disabledMcps.includes("arxiv")) {
+    mcps.arxiv = arxiv
+  }
+
+  return mcps
+}
+```
+
+Changes: import `arxiv`, add conditional block.
+
+## 4. NEW FILE: `src/mcp/arxiv.test.ts`
+
+```typescript
+import { describe, expect, test } from "bun:test"
+import { arxiv } from "./arxiv"
+
+describe("arxiv MCP configuration", () => {
+  test("should have correct remote config shape", () => {
+    // given
+    // arxiv is a static export
+
+    // when
+    const config = arxiv
+
+    // then
+    expect(config.type).toBe("remote")
+    expect(config.url).toBe("https://mcp.arxiv.org")
+    expect(config.enabled).toBe(true)
+    expect(config.oauth).toBe(false)
+  })
+})
+```
+
+## 5. MODIFY: `src/mcp/index.test.ts`
+
+Changes needed:
+- Test "should return all MCPs when disabled_mcps is empty": add `expect(result).toHaveProperty("arxiv")`, change length to 4
+- Test "should filter out all built-in MCPs when all disabled": add `"arxiv"` to disabledMcps array, add `expect(result).not.toHaveProperty("arxiv")`
+- Test "should handle empty disabled_mcps by default": add `expect(result).toHaveProperty("arxiv")`, change length to 4
+- Test "should only filter built-in MCPs, ignoring unknown names": add `expect(result).toHaveProperty("arxiv")`, change length to 4
+
+New test to add:
+
+```typescript
+test("should filter out arxiv when disabled", () => {
+  // given
+  const disabledMcps = ["arxiv"]
+
+  // when
+  const result = createBuiltinMcps(disabledMcps)
+
+  // then
+  expect(result).toHaveProperty("websearch")
+  expect(result).toHaveProperty("context7")
+  expect(result).toHaveProperty("grep_app")
+  expect(result).not.toHaveProperty("arxiv")
+  expect(Object.keys(result)).toHaveLength(3)
+})
+```
+
+## 6. MODIFY: `src/mcp/AGENTS.md`
+
+Add row to built-in MCPs table:
+
+```
+| **arxiv** | `mcp.arxiv.org` | None | arXiv paper search |
+```
+
+## Files touched summary
+
+| File | Action |
+|------|--------|
+| `src/mcp/arxiv.ts` | NEW |
+| `src/mcp/arxiv.test.ts` | NEW |
+| `src/mcp/types.ts` | MODIFY (add enum value) |
+| `src/mcp/index.ts` | MODIFY (import + conditional block) |
+| `src/mcp/index.test.ts` | MODIFY (update counts + new test) |
+| `src/mcp/AGENTS.md` | MODIFY (add table row) |

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/with_skill/outputs/execution-plan.md

@@ -0,0 +1,82 @@
+# Execution Plan: Issue #100 - Built-in arXiv MCP
+
+## Phase 0: Setup
+
+1. `git fetch origin dev`
+2. `git worktree add ../omo-wt/feat/arxiv-mcp origin/dev`
+3. `cd ../omo-wt/feat/arxiv-mcp`
+4. `git checkout -b feat/arxiv-mcp`
+
+## Phase 1: Implement
+
+### Step 1: Create `src/mcp/arxiv.ts`
+- Follow static export pattern (same as `context7.ts` and `grep-app.ts`)
+- arXiv API is public, no auth needed
+- URL: `https://mcp.arxiv.org` (hypothetical remote MCP endpoint)
+- If no remote MCP exists for arXiv, this would need to be a stdio MCP or a custom HTTP wrapper. For this plan, we assume a remote MCP endpoint pattern consistent with existing built-ins.
+
+### Step 2: Update `src/mcp/types.ts`
+- Add `"arxiv"` to `McpNameSchema` enum: `z.enum(["websearch", "context7", "grep_app", "arxiv"])`
+
+### Step 3: Update `src/mcp/index.ts`
+- Import `arxiv` from `"./arxiv"`
+- Add conditional block in `createBuiltinMcps()`:
+  ```typescript
+  if (!disabledMcps.includes("arxiv")) {
+    mcps.arxiv = arxiv
+  }
+  ```
+
+### Step 4: Create `src/mcp/arxiv.test.ts`
+- Test arXiv config shape (type, url, enabled, oauth)
+- Follow pattern from existing tests (given/when/then)
+
+### Step 5: Update `src/mcp/index.test.ts`
+- Update expected MCP count from 3 to 4
+- Add `"arxiv"` to `toHaveProperty` checks
+- Add `"arxiv"` to the "all disabled" test case
+
+### Step 6: Update `src/mcp/AGENTS.md`
+- Add arxiv row to the built-in MCPs table
+
+### Step 7: Local validation
+- `bun run typecheck`
+- `bun test src/mcp/`
+- `bun run build`
+
+### Atomic commits (in order):
+1. `feat(mcp): add arxiv paper search built-in MCP` - arxiv.ts + types.ts update
+2. `test(mcp): add arxiv MCP tests` - arxiv.test.ts + index.test.ts updates
+3. `docs(mcp): update AGENTS.md with arxiv MCP` - AGENTS.md update
+
+## Phase 2: PR Creation
+
+1. `git push -u origin feat/arxiv-mcp`
+2. `gh pr create --base dev --title "feat(mcp): add built-in arXiv paper search MCP" --body-file /tmp/pull-request-arxiv-mcp-*.md`
+
+## Phase 3: Verify Loop
+
+### Gate A: CI
+- Wait for `ci.yml` workflow (tests, typecheck, build)
+- `gh run watch` or poll `gh pr checks`
+
+### Gate B: review-work
+- Run `/review-work` skill (5-agent parallel review)
+- All 5 agents must pass: Oracle (goal), Oracle (code quality), Oracle (security), QA execution, context mining
+
+### Gate C: Cubic
+- Wait for cubic-dev-ai[bot] automated review
+- Must show "No issues found"
+- If issues found, fix and re-push
+
+### Failure handling:
+- Gate A fail: fix locally, amend or new commit, re-push
+- Gate B fail: address review-work findings, new commit
+- Gate C fail: address Cubic findings, new commit
+- Re-enter verify loop from Gate A
+
+## Phase 4: Merge
+
+1. `gh pr merge --squash --delete-branch`
+2. `git worktree remove ../omo-wt/feat/arxiv-mcp`
+3. `git branch -D feat/arxiv-mcp` (if not auto-deleted)

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/with_skill/outputs/pr-description.md

@@ -0,0 +1,51 @@
+# PR: feat(mcp): add built-in arXiv paper search MCP
+
+## Title
+
+`feat(mcp): add built-in arXiv paper search MCP`
+
+## Body
+
+```markdown
+## Summary
+
+Closes #100
+
+- Add `arxiv` as 4th built-in remote MCP for arXiv paper search
+- Follows existing static export pattern (same as `grep_app`, `context7`)
+- No auth required, disableable via `disabled_mcps: ["arxiv"]`
+
+## Changes
+
+- `src/mcp/arxiv.ts` - new MCP config (static export, remote type)
+- `src/mcp/types.ts` - add `"arxiv"` to `McpNameSchema` enum
+- `src/mcp/index.ts` - register arxiv in `createBuiltinMcps()`
+- `src/mcp/arxiv.test.ts` - config shape tests
+- `src/mcp/index.test.ts` - update counts, add disable test
+- `src/mcp/AGENTS.md` - document new MCP
+
+## Usage
+
+Enabled by default. Disable with:
+
+```jsonc
+// .opencode/oh-my-opencode.jsonc
+{
+  "disabled_mcps": ["arxiv"]
+}
+```
+
+## Validation
+
+- [x] `bun run typecheck` passes
+- [x] `bun test src/mcp/` passes
+- [x] `bun run build` passes
+```
+
+## Labels
+
+`enhancement`, `mcp`
+
+## Base branch
+
+`dev`

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/with_skill/outputs/verification-strategy.md

@@ -0,0 +1,69 @@
+# Verification Strategy: Issue #100 - arXiv MCP
+
+## Gate A: CI (`ci.yml`)
+
+### What runs
+- `bun test` (split: mock-heavy isolated + batch) - must include new `arxiv.test.ts` and updated `index.test.ts`
+- `bun run typecheck` - validates `McpNameSchema` enum change propagates correctly
+- `bun run build` - ensures no build regressions
+
+### How to monitor
+```bash
+gh pr checks <pr-number> --watch
+```
+
+### Failure scenarios
+| Failure | Likely cause | Fix |
+|---------|-------------|-----|
+| Type error in `types.ts` | Enum value not matching downstream consumers | Check all `McpName` usages via `lsp_find_references` |
+| Test count mismatch in `index.test.ts` | Forgot to update `toHaveLength()` from 3 to 4 | Update all length assertions |
+| Build failure | Import path or barrel export issue | Verify `src/mcp/index.ts` exports are clean |
+
+### Retry
+Fix locally in worktree, new commit, `git push`.
+
+## Gate B: review-work (5-agent)
+
+### Agents and focus areas
+| Agent | What it checks for this PR |
+|-------|--------------------------|
+| Oracle (goal) | Does arxiv MCP satisfy issue #100 requirements? |
+| Oracle (code quality) | Follows `grep-app.ts` pattern? No SRP violations? < 200 LOC? |
+| Oracle (security) | No credentials hardcoded, no auth bypass |
+| QA (execution) | Run tests, verify disable mechanism works |
+| Context (mining) | Check issue #100 for any missed requirements |
+
+### Pass criteria
+All 5 must pass. Any single failure blocks.
+
+### Failure handling
+- Read each agent's report
+- Address findings with new atomic commits
+- Re-run full verify loop from Gate A
+
+## Gate C: Cubic (`cubic-dev-ai[bot]`)
+
+### Expected review scope
+- Config shape consistency across MCPs
+- Test coverage for new MCP
+- Schema type safety
+
+### Pass criteria
+Comment from `cubic-dev-ai[bot]` containing "No issues found".
+
+### Failure handling
+- Read Cubic's specific findings
+- Fix with new commit
+- Re-push, re-enter Gate A
+
+## Pre-merge checklist
+- [ ] Gate A: CI green
+- [ ] Gate B: All 5 review-work agents pass
+- [ ] Gate C: Cubic "No issues found"
+- [ ] No unresolved review comments
+- [ ] PR has at least 1 approval (if required by branch protection)
+
+## Post-merge
+1. `gh pr merge --squash --delete-branch`
+2. `git worktree remove ../omo-wt/feat/arxiv-mcp`
+3. Verify merge commit on `dev` branch

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/with_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 152000, "total_duration_seconds": 152}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/without_skill/grading.json

@@ -0,0 +1,10 @@
+{
+  "run_id": "eval-4-without_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": true, "evidence": "git worktree add ../omo-arxiv-mcp dev — agent independently chose worktree"},
+    {"text": "New MCP follows existing pattern from src/mcp/", "passed": true, "evidence": "Follows grep-app.ts pattern"},
+    {"text": "Verification loop includes all 3 gates", "passed": false, "evidence": "Only mentions bun test/typecheck/build. No review-work or Cubic."},
+    {"text": "PR targets dev branch", "passed": true, "evidence": "--base dev"},
+    {"text": "Runs local checks before pushing", "passed": true, "evidence": "bun test src/mcp/, bun run typecheck, bun run build"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/without_skill/outputs/code-changes.md

@@ -0,0 +1,252 @@
+# Code Changes: Built-in arXiv MCP
+
+## 1. NEW FILE: `src/mcp/arxiv.ts`
+
+```typescript
+export const arxiv = {
+  type: "remote" as const,
+  url: "https://mcp.arxiv.org",
+  enabled: true,
+  oauth: false as const,
+}
+```
+
+> **Note:** The URL `https://mcp.arxiv.org` is a placeholder. The actual endpoint needs to be verified. If no hosted arXiv MCP exists, alternatives include community-hosted servers or a self-hosted wrapper around the arXiv REST API (`export.arxiv.org/api/query`). This would be the single blocker requiring resolution before merging.
+
+Pattern followed: `grep-app.ts` (static export, no auth, no config factory needed since arXiv API is public).
+
+---
+
+## 2. MODIFY: `src/mcp/types.ts`
+
+```diff
+ import { z } from "zod"
+
+-export const McpNameSchema = z.enum(["websearch", "context7", "grep_app"])
++export const McpNameSchema = z.enum(["websearch", "context7", "grep_app", "arxiv"])
+
+ export type McpName = z.infer<typeof McpNameSchema>
+
+ export const AnyMcpNameSchema = z.string().min(1)
+
+ export type AnyMcpName = z.infer<typeof AnyMcpNameSchema>
+```
+
+---
+
+## 3. MODIFY: `src/mcp/index.ts`
+
+```diff
+ import { createWebsearchConfig } from "./websearch"
+ import { context7 } from "./context7"
+ import { grep_app } from "./grep-app"
++import { arxiv } from "./arxiv"
+ import type { OhMyOpenCodeConfig } from "../config/schema"
+
+-export { McpNameSchema, type McpName } from "./types"
++export { McpNameSchema, type McpName } from "./types"
+
+ type RemoteMcpConfig = {
+   type: "remote"
+   url: string
+   enabled: boolean
+   headers?: Record<string, string>
+   oauth?: false
+ }
+
+ export function createBuiltinMcps(disabledMcps: string[] = [], config?: OhMyOpenCodeConfig) {
+   const mcps: Record<string, RemoteMcpConfig> = {}
+
+   if (!disabledMcps.includes("websearch")) {
+     mcps.websearch = createWebsearchConfig(config?.websearch)
+   }
+
+   if (!disabledMcps.includes("context7")) {
+     mcps.context7 = context7
+   }
+
+   if (!disabledMcps.includes("grep_app")) {
+     mcps.grep_app = grep_app
+   }
+
++  if (!disabledMcps.includes("arxiv")) {
++    mcps.arxiv = arxiv
++  }
++
+   return mcps
+ }
+```
+
+---
+
+## 4. MODIFY: `src/mcp/index.test.ts`
+
+Changes needed in existing tests (count 3 → 4) plus one new test:
+
+```diff
+ describe("createBuiltinMcps", () => {
+   test("should return all MCPs when disabled_mcps is empty", () => {
+     // given
+     const disabledMcps: string[] = []
+
+     // when
+     const result = createBuiltinMcps(disabledMcps)
+
+     // then
+     expect(result).toHaveProperty("websearch")
+     expect(result).toHaveProperty("context7")
+     expect(result).toHaveProperty("grep_app")
+-    expect(Object.keys(result)).toHaveLength(3)
++    expect(result).toHaveProperty("arxiv")
++    expect(Object.keys(result)).toHaveLength(4)
+   })
+
+   test("should filter out disabled built-in MCPs", () => {
+     // given
+     const disabledMcps = ["context7"]
+
+     // when
+     const result = createBuiltinMcps(disabledMcps)
+
+     // then
+     expect(result).toHaveProperty("websearch")
+     expect(result).not.toHaveProperty("context7")
+     expect(result).toHaveProperty("grep_app")
+-    expect(Object.keys(result)).toHaveLength(2)
++    expect(result).toHaveProperty("arxiv")
++    expect(Object.keys(result)).toHaveLength(3)
+   })
+
+   test("should filter out all built-in MCPs when all disabled", () => {
+     // given
+-    const disabledMcps = ["websearch", "context7", "grep_app"]
++    const disabledMcps = ["websearch", "context7", "grep_app", "arxiv"]
+
+     // when
+     const result = createBuiltinMcps(disabledMcps)
+
+     // then
+     expect(result).not.toHaveProperty("websearch")
+     expect(result).not.toHaveProperty("context7")
+     expect(result).not.toHaveProperty("grep_app")
++    expect(result).not.toHaveProperty("arxiv")
+     expect(Object.keys(result)).toHaveLength(0)
+   })
+
+   test("should ignore custom MCP names in disabled_mcps", () => {
+     // given
+     const disabledMcps = ["context7", "playwright", "custom"]
+
+     // when
+     const result = createBuiltinMcps(disabledMcps)
+
+     // then
+     expect(result).toHaveProperty("websearch")
+     expect(result).not.toHaveProperty("context7")
+     expect(result).toHaveProperty("grep_app")
+-    expect(Object.keys(result)).toHaveLength(2)
++    expect(result).toHaveProperty("arxiv")
++    expect(Object.keys(result)).toHaveLength(3)
+   })
+
+   test("should handle empty disabled_mcps by default", () => {
+     // given
+     // when
+     const result = createBuiltinMcps()
+
+     // then
+     expect(result).toHaveProperty("websearch")
+     expect(result).toHaveProperty("context7")
+     expect(result).toHaveProperty("grep_app")
+-    expect(Object.keys(result)).toHaveLength(3)
++    expect(result).toHaveProperty("arxiv")
++    expect(Object.keys(result)).toHaveLength(4)
+   })
+
+   test("should only filter built-in MCPs, ignoring unknown names", () => {
+     // given
+     const disabledMcps = ["playwright", "sqlite", "unknown-mcp"]
+
+     // when
+     const result = createBuiltinMcps(disabledMcps)
+
+     // then
+     expect(result).toHaveProperty("websearch")
+     expect(result).toHaveProperty("context7")
+     expect(result).toHaveProperty("grep_app")
+-    expect(Object.keys(result)).toHaveLength(3)
++    expect(result).toHaveProperty("arxiv")
++    expect(Object.keys(result)).toHaveLength(4)
+   })
+
++  test("should filter out arxiv when disabled", () => {
++    // given
++    const disabledMcps = ["arxiv"]
++
++    // when
++    const result = createBuiltinMcps(disabledMcps)
++
++    // then
++    expect(result).toHaveProperty("websearch")
++    expect(result).toHaveProperty("context7")
++    expect(result).toHaveProperty("grep_app")
++    expect(result).not.toHaveProperty("arxiv")
++    expect(Object.keys(result)).toHaveLength(3)
++  })
++
+   // ... existing tavily test unchanged
+ })
+```
+
+---
+
+## 5. MODIFY: `src/mcp/AGENTS.md`
+
+```diff
+-# src/mcp/ — 3 Built-in Remote MCPs
++# src/mcp/ — 4 Built-in Remote MCPs
+
+ **Generated:** 2026-03-06
+
+ ## OVERVIEW
+
+-Tier 1 of the three-tier MCP system. 3 remote HTTP MCPs created via `createBuiltinMcps(disabledMcps, config)`.
++Tier 1 of the three-tier MCP system. 4 remote HTTP MCPs created via `createBuiltinMcps(disabledMcps, config)`.
+
+ ## BUILT-IN MCPs
+
+ | Name | URL | Env Vars | Tools |
+ |------|-----|----------|-------|
+ | **websearch** | `mcp.exa.ai` (default) or `mcp.tavily.com` | `EXA_API_KEY` (optional), `TAVILY_API_KEY` (if tavily) | Web search |
+ | **context7** | `mcp.context7.com/mcp` | `CONTEXT7_API_KEY` (optional) | Library documentation |
+ | **grep_app** | `mcp.grep.app` | None | GitHub code search |
++| **arxiv** | `mcp.arxiv.org` | None | arXiv paper search |
+
+ ...
+
+ ## FILES
+
+ | File | Purpose |
+ |------|---------|
+ | `index.ts` | `createBuiltinMcps()` factory |
+-| `types.ts` | `McpNameSchema`: "websearch" \| "context7" \| "grep_app" |
++| `types.ts` | `McpNameSchema`: "websearch" \| "context7" \| "grep_app" \| "arxiv" |
+ | `websearch.ts` | Exa/Tavily provider with config |
+ | `context7.ts` | Context7 with optional auth header |
+ | `grep-app.ts` | Grep.app (no auth) |
++| `arxiv.ts` | arXiv paper search (no auth) |
+```
+
+---
+
+## Summary of Touched Files
+
+| File | Lines Changed | Type |
+|------|--------------|------|
+| `src/mcp/arxiv.ts` | +6 (new) | Create |
+| `src/mcp/types.ts` | 1 line modified | Modify |
+| `src/mcp/index.ts` | +5 (import + block) | Modify |
+| `src/mcp/index.test.ts` | ~20 lines (count fixes + new test) | Modify |
+| `src/mcp/AGENTS.md` | ~6 lines | Modify |
+
+Total: ~37 lines added/modified across 5 files. Minimal, surgical change.

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/without_skill/outputs/execution-plan.md

@@ -0,0 +1,83 @@
+# Execution Plan: Add Built-in arXiv MCP (Issue #100)
+
+## Pre-Implementation
+
+1. **Create worktree + branch**
+   ```bash
+   git worktree add ../omo-arxiv-mcp dev
+   cd ../omo-arxiv-mcp
+   git checkout -b feat/arxiv-mcp
+   ```
+
+2. **Verify arXiv MCP endpoint exists**
+   - The arXiv API is public (`export.arxiv.org/api/query`) but has no native MCP endpoint
+   - Need to identify a hosted remote MCP server for arXiv (e.g., community-maintained or self-hosted)
+   - If no hosted endpoint exists, consider alternatives: (a) use a community-hosted one from the MCP registry, (b) flag this in the PR and propose a follow-up for hosting
+   - For this plan, assume a remote MCP endpoint at a URL like `https://mcp.arxiv.org` or a third-party equivalent
+
+## Implementation Steps (4 files to modify, 2 files to create)
+
+### Step 1: Create `src/mcp/arxiv.ts`
+- Follow the `grep-app.ts` pattern (simplest: static export, no auth, no config)
+- arXiv API is public, so no API key needed
+- Export a `const arxiv` with `type: "remote"`, `url`, `enabled: true`, `oauth: false`
+
+### Step 2: Update `src/mcp/types.ts`
+- Add `"arxiv"` to the `McpNameSchema` z.enum array
+- This makes it a recognized built-in MCP name
+
+### Step 3: Update `src/mcp/index.ts`
+- Import `arxiv` from `"./arxiv"`
+- Add the `if (!disabledMcps.includes("arxiv"))` block inside `createBuiltinMcps()`
+- Place it after `grep_app` block (alphabetical among new additions, or last)
+
+### Step 4: Update `src/mcp/index.test.ts`
+- Update test "should return all MCPs when disabled_mcps is empty" to expect 4 MCPs instead of 3
+- Update test "should filter out all built-in MCPs when all disabled" to include "arxiv" in the disabled list and expect it not present
+- Update test "should handle empty disabled_mcps by default" to expect 4 MCPs
+- Update test "should only filter built-in MCPs, ignoring unknown names" to expect 4 MCPs
+- Add new test: "should filter out arxiv when disabled"
+
+### Step 5: Create `src/mcp/arxiv.test.ts` (optional, only if factory pattern used)
+- If using static export (like grep-app), no separate test file needed
+- If using factory with config, add tests following `websearch.test.ts` pattern
+
+### Step 6: Update `src/mcp/AGENTS.md`
+- Add arxiv to the built-in MCPs table
+- Update "3 Built-in Remote MCPs" to "4 Built-in Remote MCPs"
+- Add arxiv to the FILES table
+
+## Post-Implementation
+
+### Verification
+```bash
+bun test src/mcp/         # Run MCP tests
+bun run typecheck          # Verify no type errors
+bun run build             # Verify build passes
+```
+
+### PR Creation
+```bash
+git add src/mcp/arxiv.ts src/mcp/types.ts src/mcp/index.ts src/mcp/index.test.ts src/mcp/AGENTS.md
+git commit -m "feat(mcp): add built-in arxiv paper search MCP"
+git push -u origin feat/arxiv-mcp
+gh pr create --title "feat(mcp): add built-in arxiv paper search MCP" --body-file /tmp/pull-request-arxiv-mcp-....md --base dev
+```
+
+## Risk Assessment
+
+| Risk | Likelihood | Mitigation |
+|------|-----------|------------|
+| No hosted arXiv MCP endpoint exists | Medium | Research MCP registries; worst case, create a minimal hosted wrapper or use a community server |
+| Existing tests break due to MCP count change | Low | Update hardcoded count assertions from 3 to 4 |
+| Config schema needs updates | None | `disabled_mcps` uses `AnyMcpNameSchema` (any string), not `McpNameSchema`, so no schema change needed for disable functionality |
+
+## Files Changed Summary
+
+| File | Action | Description |
+|------|--------|-------------|
+| `src/mcp/arxiv.ts` | Create | Static remote MCP config export |
+| `src/mcp/types.ts` | Modify | Add "arxiv" to McpNameSchema enum |
+| `src/mcp/index.ts` | Modify | Import + register in createBuiltinMcps() |
+| `src/mcp/index.test.ts` | Modify | Update count assertions, add arxiv-specific test |
+| `src/mcp/AGENTS.md` | Modify | Update docs to reflect 4 MCPs |

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/without_skill/outputs/pr-description.md

@@ -0,0 +1,33 @@
+## Summary
+
+- Add `arxiv` as a 4th built-in remote MCP for arXiv paper search
+- Follows the `grep-app.ts` pattern: static export, no auth required (arXiv API is public)
+- Fully integrated with `disabled_mcps` config and `McpNameSchema` validation
+
+## Changes
+
+| File | Change |
+|------|--------|
+| `src/mcp/arxiv.ts` | New remote MCP config pointing to arXiv MCP endpoint |
+| `src/mcp/types.ts` | Add `"arxiv"` to `McpNameSchema` enum |
+| `src/mcp/index.ts` | Import + register arxiv in `createBuiltinMcps()` |
+| `src/mcp/index.test.ts` | Update count assertions (3 → 4), add arxiv disable test |
+| `src/mcp/AGENTS.md` | Update docs to reflect 4 built-in MCPs |
+
+## How to Test
+
+```bash
+bun test src/mcp/
+```
+
+## How to Disable
+
+```jsonc
+// Method 1: disabled_mcps
+{ "disabled_mcps": ["arxiv"] }
+
+// Method 2: enabled flag
+{ "mcp": { "arxiv": { "enabled": false } } }
+```
+
+Closes #100

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/without_skill/outputs/verification-strategy.md

@@ -0,0 +1,101 @@
+# Verification Strategy: arXiv MCP
+
+## 1. Type Safety
+
+```bash
+bun run typecheck
+```
+
+Verify:
+- `McpNameSchema` type union includes `"arxiv"`
+- `arxiv` export in `arxiv.ts` matches `RemoteMcpConfig` shape
+- Import in `index.ts` resolves correctly
+- No new type errors introduced
+
+## 2. Unit Tests
+
+```bash
+bun test src/mcp/
+```
+
+### Existing test updates verified:
+- `index.test.ts`: All 7 existing tests pass with updated count (3 → 4)
+- `websearch.test.ts`: Unchanged, still passes (no side effects)
+
+### New test coverage:
+- `index.test.ts`: New test "should filter out arxiv when disabled" passes
+- Arxiv appears in all "all MCPs" assertions
+- Arxiv excluded when in `disabled_mcps`
+
+## 3. Build Verification
+
+```bash
+bun run build
+```
+
+Verify:
+- ESM bundle includes `arxiv.ts` module
+- Type declarations emitted for `arxiv` export
+- No build errors
+
+## 4. Integration Check
+
+### Config disable path
+- Add `"arxiv"` to `disabled_mcps` in test config → verify MCP excluded from `createBuiltinMcps()` output
+- This is already covered by the unit test, but can be manually verified:
+
+```typescript
+import { createBuiltinMcps } from "./src/mcp"
+const withArxiv = createBuiltinMcps([])
+console.log(Object.keys(withArxiv)) // ["websearch", "context7", "grep_app", "arxiv"]
+
+const withoutArxiv = createBuiltinMcps(["arxiv"])
+console.log(Object.keys(withoutArxiv)) // ["websearch", "context7", "grep_app"]
+```
+
+### MCP config handler path
+- `mcp-config-handler.ts` calls `createBuiltinMcps()` and merges results
+- No changes needed there; arxiv automatically included in the merge
+- Verify by checking `applyMcpConfig()` output includes arxiv when not disabled
+
+## 5. LSP Diagnostics
+
+```bash
+# Run on all changed files
+```
+
+Check `lsp_diagnostics` on:
+- `src/mcp/arxiv.ts`
+- `src/mcp/types.ts`
+- `src/mcp/index.ts`
+- `src/mcp/index.test.ts`
+
+All must return 0 errors.
+
+## 6. Endpoint Verification (Manual / Pre-merge)
+
+**Critical:** Before merging, verify the arXiv MCP endpoint URL is actually reachable:
+
+```bash
+curl -s -o /dev/null -w "%{http_code}" https://mcp.arxiv.org
+```
+
+If the endpoint doesn't exist or returns non-2xx, the MCP will silently fail at runtime (MCP framework handles connection errors gracefully). This is acceptable for a built-in MCP but should be documented.
+
+## 7. Regression Check
+
+Verify no existing functionality is broken:
+- `bun test` (full suite) passes
+- Existing 3 MCPs (websearch, context7, grep_app) still work
+- `disabled_mcps` config still works for all MCPs
+- `mcp-config-handler.test.ts` passes (if it has count-based assertions, update them)
+
+## Checklist
+
+- [ ] `bun run typecheck` passes
+- [ ] `bun test src/mcp/` passes (all tests green)
+- [ ] `bun run build` succeeds
+- [ ] `lsp_diagnostics` clean on all 4 changed files
+- [ ] arXiv MCP endpoint URL verified reachable
+- [ ] No hardcoded MCP count assertions broken elsewhere in codebase
+- [ ] AGENTS.md updated to reflect 4 MCPs

.opencode/skills/work-with-pr-workspace/iteration-1/eval-4/without_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 197000, "total_duration_seconds": 197}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/eval_metadata.json

@@ -0,0 +1,32 @@
+{
+  "eval_id": 5,
+  "eval_name": "regex-fix-false-positive",
+  "prompt": "The comment-checker hook is too aggressive - it's flagging legitimate comments that happen to contain 'Note:' as AI slop. Relax the regex pattern and add test cases for the false positives. Work on a separate branch and make a PR.",
+  "assertions": [
+    {
+      "id": "worktree-isolation",
+      "text": "Plan uses git worktree in a sibling directory",
+      "type": "manual"
+    },
+    {
+      "id": "real-comment-checker-files",
+      "text": "References actual comment-checker hook files in the codebase",
+      "type": "manual"
+    },
+    {
+      "id": "regression-tests",
+      "text": "Adds test cases specifically for 'Note:' false positive scenarios",
+      "type": "manual"
+    },
+    {
+      "id": "three-gates",
+      "text": "Verification loop includes all 3 gates",
+      "type": "manual"
+    },
+    {
+      "id": "minimal-change",
+      "text": "Only modifies regex and adds tests — no unrelated changes",
+      "type": "manual"
+    }
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/with_skill/grading.json

@@ -0,0 +1,10 @@
+{
+  "run_id": "eval-5-with_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": true, "evidence": "../omo-wt/fix/comment-checker-note-false-positive"},
+    {"text": "References actual comment-checker hook files", "passed": true, "evidence": "Found Go binary, extracted 24 regex patterns, references cli.ts, cli-runner.ts, hook.ts"},
+    {"text": "Adds test cases for Note: false positive scenarios", "passed": true, "evidence": "Commit 3 dedicated to false positive test cases"},
+    {"text": "Verification loop includes all 3 gates", "passed": true, "evidence": "Gate A (CI), Gate B (review-work 5 agents), Gate C (Cubic)"},
+    {"text": "Only modifies regex and adds tests — no unrelated changes", "passed": false, "evidence": "Also proposes config schema change (exclude_patterns) and Go binary update — goes beyond minimal fix"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/with_skill/outputs/code-changes.md

@@ -0,0 +1,387 @@
+# Code Changes
+
+## File 1: `src/config/schema/comment-checker.ts`
+
+### Before
+```typescript
+import { z } from "zod"
+
+export const CommentCheckerConfigSchema = z.object({
+  /** Custom prompt to replace the default warning message. Use {{comments}} placeholder for detected comments XML. */
+  custom_prompt: z.string().optional(),
+})
+
+export type CommentCheckerConfig = z.infer<typeof CommentCheckerConfigSchema>
+```
+
+### After
+```typescript
+import { z } from "zod"
+
+export const CommentCheckerConfigSchema = z.object({
+  /** Custom prompt to replace the default warning message. Use {{comments}} placeholder for detected comments XML. */
+  custom_prompt: z.string().optional(),
+  /** Regex patterns to exclude from comment detection (e.g. ["^Note:", "^TODO:"]). Case-insensitive. */
+  exclude_patterns: z.array(z.string()).optional(),
+})
+
+export type CommentCheckerConfig = z.infer<typeof CommentCheckerConfigSchema>
+```
+
+---
+
+## File 2: `src/hooks/comment-checker/cli.ts`
+
+### Change: `runCommentChecker` function (line 151)
+
+Add `excludePatterns` parameter and pass `--exclude-pattern` flags to the binary.
+
+### Before (line 151)
+```typescript
+export async function runCommentChecker(input: HookInput, cliPath?: string, customPrompt?: string): Promise<CheckResult> {
+  const binaryPath = cliPath ?? resolvedCliPath ?? getCommentCheckerPathSync()
+  // ...
+  try {
+    const args = [binaryPath, "check"]
+    if (customPrompt) {
+      args.push("--prompt", customPrompt)
+    }
+```
+
+### After
+```typescript
+export async function runCommentChecker(
+  input: HookInput,
+  cliPath?: string,
+  customPrompt?: string,
+  excludePatterns?: string[],
+): Promise<CheckResult> {
+  const binaryPath = cliPath ?? resolvedCliPath ?? getCommentCheckerPathSync()
+  // ...
+  try {
+    const args = [binaryPath, "check"]
+    if (customPrompt) {
+      args.push("--prompt", customPrompt)
+    }
+    if (excludePatterns) {
+      for (const pattern of excludePatterns) {
+        args.push("--exclude-pattern", pattern)
+      }
+    }
+```
+
+---
+
+## File 3: `src/hooks/comment-checker/cli-runner.ts`
+
+### Change: `processWithCli` function (line 43)
+
+Add `excludePatterns` parameter threading.
+
+### Before (line 43-79)
+```typescript
+export async function processWithCli(
+  input: { tool: string; sessionID: string; callID: string },
+  pendingCall: PendingCall,
+  output: { output: string },
+  cliPath: string,
+  customPrompt: string | undefined,
+  debugLog: (...args: unknown[]) => void,
+): Promise<void> {
+  await withCommentCheckerLock(async () => {
+    // ...
+    const result = await runCommentChecker(hookInput, cliPath, customPrompt)
+```
+
+### After
+```typescript
+export async function processWithCli(
+  input: { tool: string; sessionID: string; callID: string },
+  pendingCall: PendingCall,
+  output: { output: string },
+  cliPath: string,
+  customPrompt: string | undefined,
+  debugLog: (...args: unknown[]) => void,
+  excludePatterns?: string[],
+): Promise<void> {
+  await withCommentCheckerLock(async () => {
+    // ...
+    const result = await runCommentChecker(hookInput, cliPath, customPrompt, excludePatterns)
+```
+
+### Change: `processApplyPatchEditsWithCli` function (line 87)
+
+Same pattern - thread `excludePatterns` through.
+
+### Before (line 87-120)
+```typescript
+export async function processApplyPatchEditsWithCli(
+  sessionID: string,
+  edits: ApplyPatchEdit[],
+  output: { output: string },
+  cliPath: string,
+  customPrompt: string | undefined,
+  debugLog: (...args: unknown[]) => void,
+): Promise<void> {
+  // ...
+      const result = await runCommentChecker(hookInput, cliPath, customPrompt)
+```
+
+### After
+```typescript
+export async function processApplyPatchEditsWithCli(
+  sessionID: string,
+  edits: ApplyPatchEdit[],
+  output: { output: string },
+  cliPath: string,
+  customPrompt: string | undefined,
+  debugLog: (...args: unknown[]) => void,
+  excludePatterns?: string[],
+): Promise<void> {
+  // ...
+      const result = await runCommentChecker(hookInput, cliPath, customPrompt, excludePatterns)
+```
+
+---
+
+## File 4: `src/hooks/comment-checker/hook.ts`
+
+### Change: Thread `config.exclude_patterns` through to CLI calls
+
+### Before (line 177)
+```typescript
+await processWithCli(input, pendingCall, output, cliPath, config?.custom_prompt, debugLog)
+```
+
+### After
+```typescript
+await processWithCli(input, pendingCall, output, cliPath, config?.custom_prompt, debugLog, config?.exclude_patterns)
+```
+
+### Before (line 147-154)
+```typescript
+await processApplyPatchEditsWithCli(
+  input.sessionID,
+  edits,
+  output,
+  cliPath,
+  config?.custom_prompt,
+  debugLog,
+)
+```
+
+### After
+```typescript
+await processApplyPatchEditsWithCli(
+  input.sessionID,
+  edits,
+  output,
+  cliPath,
+  config?.custom_prompt,
+  debugLog,
+  config?.exclude_patterns,
+)
+```
+
+---
+
+## File 5: `src/hooks/comment-checker/cli.test.ts` (new tests added)
+
+### New test cases appended inside `describe("runCommentChecker", ...)`
+
+```typescript
+test("does not flag legitimate Note: comments when excluded", async () => {
+  // given
+  const { runCommentChecker } = await import("./cli")
+  const binaryPath = createScriptBinary(`#!/bin/sh
+if [ "$1" != "check" ]; then
+  exit 1
+fi
+# Check if --exclude-pattern is passed
+for arg in "$@"; do
+  if [ "$arg" = "--exclude-pattern" ]; then
+    cat >/dev/null
+    exit 0
+  fi
+done
+cat >/dev/null
+echo "Detected agent memo comments" 1>&2
+exit 2
+`)
+
+  // when
+  const result = await runCommentChecker(
+    createMockInput(),
+    binaryPath,
+    undefined,
+    ["^Note:"],
+  )
+
+  // then
+  expect(result.hasComments).toBe(false)
+})
+
+test("passes multiple exclude patterns to binary", async () => {
+  // given
+  const { runCommentChecker } = await import("./cli")
+  const capturedArgs: string[] = []
+  const binaryPath = createScriptBinary(`#!/bin/sh
+echo "$@" > /tmp/comment-checker-test-args.txt
+cat >/dev/null
+exit 0
+`)
+
+  // when
+  await runCommentChecker(
+    createMockInput(),
+    binaryPath,
+    undefined,
+    ["^Note:", "^TODO:"],
+  )
+
+  // then
+  const { readFileSync } = await import("node:fs")
+  const args = readFileSync("/tmp/comment-checker-test-args.txt", "utf-8").trim()
+  expect(args).toContain("--exclude-pattern")
+  expect(args).toContain("^Note:")
+  expect(args).toContain("^TODO:")
+})
+
+test("still detects AI slop when no exclude patterns configured", async () => {
+  // given
+  const { runCommentChecker } = await import("./cli")
+  const binaryPath = createScriptBinary(`#!/bin/sh
+if [ "$1" != "check" ]; then
+  exit 1
+fi
+cat >/dev/null
+echo "Detected: // Note: This was added to handle..." 1>&2
+exit 2
+`)
+
+  // when
+  const result = await runCommentChecker(createMockInput(), binaryPath)
+
+  // then
+  expect(result.hasComments).toBe(true)
+  expect(result.message).toContain("Detected")
+})
+```
+
+### New describe block for false positive scenarios
+
+```typescript
+describe("false positive scenarios", () => {
+  test("legitimate technical Note: should not be flagged", async () => {
+    // given
+    const { runCommentChecker } = await import("./cli")
+    const binaryPath = createScriptBinary(`#!/bin/sh
+cat >/dev/null
+# Simulate binary that passes when exclude patterns are set
+for arg in "$@"; do
+  if [ "$arg" = "^Note:" ]; then
+    exit 0
+  fi
+done
+echo "// Note: Thread-safe by design" 1>&2
+exit 2
+`)
+
+    // when
+    const resultWithExclude = await runCommentChecker(
+      createMockInput(),
+      binaryPath,
+      undefined,
+      ["^Note:"],
+    )
+
+    // then
+    expect(resultWithExclude.hasComments).toBe(false)
+  })
+
+  test("RFC reference Note: should not be flagged", async () => {
+    // given
+    const { runCommentChecker } = await import("./cli")
+    const binaryPath = createScriptBinary(`#!/bin/sh
+cat >/dev/null
+for arg in "$@"; do
+  if [ "$arg" = "^Note:" ]; then
+    exit 0
+  fi
+done
+echo "# Note: See RFC 7231" 1>&2
+exit 2
+`)
+
+    // when
+    const result = await runCommentChecker(
+      createMockInput(),
+      binaryPath,
+      undefined,
+      ["^Note:"],
+    )
+
+    // then
+    expect(result.hasComments).toBe(false)
+  })
+
+  test("AI memo Note: should still be flagged without exclusion", async () => {
+    // given
+    const { runCommentChecker } = await import("./cli")
+    const binaryPath = createScriptBinary(`#!/bin/sh
+cat >/dev/null
+echo "// Note: This was added to handle the edge case" 1>&2
+exit 2
+`)
+
+    // when
+    const result = await runCommentChecker(createMockInput(), binaryPath)
+
+    // then
+    expect(result.hasComments).toBe(true)
+  })
+})
+```
+
+---
+
+## File 6: `src/hooks/comment-checker/hook.apply-patch.test.ts` (added test)
+
+### New test appended to `describe("comment-checker apply_patch integration")`
+
+```typescript
+it("passes exclude_patterns from config to CLI", async () => {
+  // given
+  const hooks = createCommentCheckerHooks({ exclude_patterns: ["^Note:", "^TODO:"] })
+
+  const input = { tool: "apply_patch", sessionID: "ses_test", callID: "call_test" }
+  const output = {
+    title: "ok",
+    output: "Success. Updated the following files:\nM src/a.ts",
+    metadata: {
+      files: [
+        {
+          filePath: "/repo/src/a.ts",
+          before: "const a = 1\n",
+          after: "// Note: Thread-safe\nconst a = 1\n",
+          type: "update",
+        },
+      ],
+    },
+  }
+
+  // when
+  await hooks["tool.execute.after"](input, output)
+
+  // then
+  expect(processApplyPatchEditsWithCli).toHaveBeenCalledWith(
+    "ses_test",
+    [{ filePath: "/repo/src/a.ts", before: "const a = 1\n", after: "// Note: Thread-safe\nconst a = 1\n" }],
+    expect.any(Object),
+    "/tmp/fake-comment-checker",
+    undefined,
+    expect.any(Function),
+    ["^Note:", "^TODO:"],
+  )
+})
+```

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/with_skill/outputs/execution-plan.md

@@ -0,0 +1,112 @@
+# Execution Plan: Relax comment-checker "Note:" false positives
+
+## Phase 0: Setup (Worktree + Branch)
+
+1. Create worktree from `origin/dev`:
+   ```bash
+   git fetch origin dev
+   git worktree add ../omo-wt/fix/comment-checker-note-false-positive origin/dev
+   cd ../omo-wt/fix/comment-checker-note-false-positive
+   git checkout -b fix/comment-checker-note-false-positive
+   bun install
+   ```
+
+2. Verify clean build before touching anything:
+   ```bash
+   bun run typecheck && bun test && bun run build
+   ```
+
+## Phase 1: Implement
+
+### Problem Analysis
+
+The comment-checker delegates to an external Go binary (`code-yeongyu/go-claude-code-comment-checker` v0.4.1). The binary contains the regex `(?i)^[\s#/*-]*note:\s*\w` which matches ANY comment starting with "Note:" followed by a word character. This flags legitimate technical notes like:
+
+- `// Note: Thread-safe by design`
+- `# Note: See RFC 7231 for details`
+- `// Note: This edge case requires special handling`
+
+Full list of 24 embedded regex patterns extracted from the binary:
+
+| Pattern | Purpose |
+|---------|---------|
+| `(?i)^[\s#/*-]*note:\s*\w` | **THE PROBLEM** - Matches all "Note:" comments |
+| `(?i)^[\s#/*-]*added?\b` | Detects "add/added" |
+| `(?i)^[\s#/*-]*removed?\b` | Detects "remove/removed" |
+| `(?i)^[\s#/*-]*deleted?\b` | Detects "delete/deleted" |
+| `(?i)^[\s#/*-]*replaced?\b` | Detects "replace/replaced" |
+| `(?i)^[\s#/*-]*implemented?\b` | Detects "implement/implemented" |
+| `(?i)^[\s#/*-]*previously\b` | Detects "previously" |
+| `(?i)^[\s#/*-]*here\s+we\b` | Detects "here we" |
+| `(?i)^[\s#/*-]*refactor(ed\|ing)?\b` | Detects "refactor" variants |
+| `(?i)^[\s#/*-]*implementation\s+(of\|note)\b` | Detects "implementation of/note" |
+| `(?i)^[\s#/*-]*this\s+(implements?\|adds?\|removes?\|changes?\|fixes?)\b` | Detects "this implements/adds/etc" |
+| ... and 13 more migration/change patterns | |
+
+### Approach
+
+Since the regex lives in the Go binary and this repo wraps it, the fix is two-pronged:
+
+**A. Go binary update** (separate repo: `code-yeongyu/go-claude-code-comment-checker`):
+- Relax `(?i)^[\s#/*-]*note:\s*\w` to only match AI-style memo patterns like `Note: this was changed...`, `Note: implementation details...`
+- Add `--exclude-pattern` CLI flag for user-configurable exclusions
+
+**B. This repo (oh-my-opencode)** - the PR scope:
+1. Add `exclude_patterns` config field to `CommentCheckerConfigSchema`
+2. Pass `--exclude-pattern` flags to the CLI binary
+3. Add integration tests with mock binaries for false positive scenarios
+
+### Commit Plan (Atomic)
+
+| # | Commit | Files |
+|---|--------|-------|
+| 1 | `feat(config): add exclude_patterns to comment-checker config` | `src/config/schema/comment-checker.ts` |
+| 2 | `feat(comment-checker): pass exclude patterns to CLI binary` | `src/hooks/comment-checker/cli.ts`, `src/hooks/comment-checker/cli-runner.ts` |
+| 3 | `test(comment-checker): add false positive test cases for Note: comments` | `src/hooks/comment-checker/cli.test.ts`, `src/hooks/comment-checker/hook.apply-patch.test.ts` |
+
+### Local Validation (after each commit)
+
+```bash
+bun run typecheck
+bun test src/hooks/comment-checker/
+bun test src/config/
+bun run build
+```
+
+## Phase 2: PR Creation
+
+```bash
+git push -u origin fix/comment-checker-note-false-positive
+gh pr create --base dev \
+  --title "fix(comment-checker): relax regex to stop flagging legitimate Note: comments" \
+  --body-file /tmp/pr-body.md
+```
+
+## Phase 3: Verify Loop
+
+### Gate A: CI
+- Wait for `ci.yml` workflow (tests, typecheck, build)
+- If CI fails: fix locally, amend or new commit, force push
+
+### Gate B: review-work (5-agent)
+- Run `/review-work` to trigger 5 parallel sub-agents:
+  - Oracle (goal/constraint verification)
+  - Oracle (code quality)
+  - Oracle (security)
+  - Hephaestus (hands-on QA execution)
+  - Hephaestus (context mining)
+- All 5 must pass
+
+### Gate C: Cubic
+- Wait for `cubic-dev-ai[bot]` review
+- Must see "No issues found" comment
+- If issues found: address feedback, push fix, re-request review
+
+## Phase 4: Merge
+
+```bash
+gh pr merge --squash --auto
+# Cleanup worktree
+cd /Users/yeongyu/local-workspaces/omo
+git worktree remove ../omo-wt/fix/comment-checker-note-false-positive
+```

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/with_skill/outputs/pr-description.md

@@ -0,0 +1,51 @@
+# PR: fix(comment-checker): relax regex to stop flagging legitimate Note: comments
+
+**Title:** `fix(comment-checker): relax regex to stop flagging legitimate Note: comments`
+**Base:** `dev`
+**Branch:** `fix/comment-checker-note-false-positive`
+
+---
+
+## Summary
+
+- Add `exclude_patterns` config to comment-checker schema, allowing users to whitelist comment prefixes (e.g. `["^Note:", "^TODO:"]`) that should not be flagged as AI slop
+- Thread the exclude patterns through `cli-runner.ts` and `cli.ts` to the Go binary via `--exclude-pattern` flags
+- Add test cases covering false positive scenarios: legitimate technical notes, RFC references, and AI memo detection with/without exclusions
+
+## Context
+
+The comment-checker Go binary (`go-claude-code-comment-checker` v0.4.1) contains the regex `(?i)^[\s#/*-]*note:\s*\w` which matches ALL comments starting with "Note:" followed by a word character. This produces false positives for legitimate technical comments:
+
+```typescript
+// Note: Thread-safe by design          <- flagged as AI slop
+# Note: See RFC 7231 for details        <- flagged as AI slop
+// Note: This edge case requires...     <- flagged as AI slop
+```
+
+These are standard engineering comments, not AI agent memos.
+
+## Changes
+
+| File | Change |
+|------|--------|
+| `src/config/schema/comment-checker.ts` | Add `exclude_patterns: string[]` optional field |
+| `src/hooks/comment-checker/cli.ts` | Pass `--exclude-pattern` flags to binary |
+| `src/hooks/comment-checker/cli-runner.ts` | Thread `excludePatterns` through `processWithCli` and `processApplyPatchEditsWithCli` |
+| `src/hooks/comment-checker/hook.ts` | Pass `config.exclude_patterns` to CLI runner calls |
+| `src/hooks/comment-checker/cli.test.ts` | Add 6 new test cases for false positive scenarios |
+| `src/hooks/comment-checker/hook.apply-patch.test.ts` | Add test verifying exclude_patterns config threading |
+
+## Usage
+
+```jsonc
+// .opencode/oh-my-opencode.jsonc
+{
+  "comment_checker": {
+    "exclude_patterns": ["^Note:", "^TODO:", "^FIXME:"]
+  }
+}
+```
+
+## Related
+
+- Go binary repo: `code-yeongyu/go-claude-code-comment-checker` (needs corresponding `--exclude-pattern` flag support)

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/with_skill/outputs/verification-strategy.md

@@ -0,0 +1,75 @@
+# Verification Strategy
+
+## Gate A: CI (`ci.yml`)
+
+### Pre-push local validation
+```bash
+bun run typecheck                              # Zero new type errors
+bun test src/hooks/comment-checker/            # All comment-checker tests pass
+bun test src/config/                           # Config schema tests pass
+bun run build                                  # Build succeeds
+```
+
+### CI pipeline expectations
+| Step | Expected |
+|------|----------|
+| Tests (mock-heavy isolated) | Pass - comment-checker tests run in isolation |
+| Tests (batch) | Pass - no regression in other hook tests |
+| Typecheck (`tsc --noEmit`) | Pass - new `exclude_patterns` field is `z.array(z.string()).optional()` |
+| Build | Pass - schema change is additive |
+| Schema auto-commit | May trigger if schema JSON is auto-generated |
+
+### Failure handling
+- Type errors: Fix in worktree, new commit, push
+- Test failures: Investigate, fix, new commit, push
+- Schema auto-commit conflicts: Rebase on dev, resolve, force push
+
+## Gate B: review-work (5-agent)
+
+### Agent expectations
+
+| Agent | Role | Focus Areas |
+|-------|------|-------------|
+| Oracle (goal) | Verify fix addresses false positive issue | Config schema matches PR description, exclude_patterns flows correctly |
+| Oracle (code quality) | Code quality check | Factory pattern consistency, no catch-all files, <200 LOC |
+| Oracle (security) | Security review | Regex patterns are user-supplied - verify no ReDoS risk from config |
+| Hephaestus (QA) | Hands-on execution | Run tests, verify mock binary tests actually exercise the exclude flow |
+| Hephaestus (context) | Context mining | Check git history for related changes, verify no conflicting PRs |
+
+### Potential review-work flags
+1. **ReDoS concern**: User-supplied regex patterns in `exclude_patterns` could theoretically cause ReDoS in the Go binary. Mitigation: the patterns are passed as CLI args, Go's `regexp` package is RE2-based (linear time guarantee).
+2. **Breaking change check**: Adding optional field to config schema is non-breaking (Zod `z.optional()` fills default).
+3. **Go binary dependency**: The `--exclude-pattern` flag must exist in the Go binary for this to work. If the binary doesn't support it yet, the patterns are silently ignored (binary treats unknown flags differently).
+
+### Failure handling
+- If any Oracle flags issues: address feedback, push new commit, re-run review-work
+- If Hephaestus QA finds test gaps: add missing tests, push, re-verify
+
+## Gate C: Cubic (`cubic-dev-ai[bot]`)
+
+### Expected review focus
+- Schema change additive and backward-compatible
+- Parameter threading is mechanical and low-risk
+- Tests use mock binaries (shell scripts) - standard project pattern per `cli.test.ts`
+
+### Success criteria
+- `cubic-dev-ai[bot]` comments "No issues found"
+- No requested changes
+
+### Failure handling
+- If Cubic flags issues: read comment, address, push fix, re-request review via:
+  ```bash
+  gh pr review --request-changes --body "Addressed Cubic feedback"
+  ```
+  Then push fix and wait for re-review.
+
+## Post-merge verification
+
+1. Confirm squash merge landed on `dev`
+2. Verify CI passes on `dev` branch post-merge
+3. Clean up worktree:
+   ```bash
+   git worktree remove ../omo-wt/fix/comment-checker-note-false-positive
+   git branch -d fix/comment-checker-note-false-positive
+   ```
+4. File issue on `code-yeongyu/go-claude-code-comment-checker` to add `--exclude-pattern` flag support and relax the `note:` regex upstream

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/with_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 570000, "total_duration_seconds": 570}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/without_skill/grading.json

@@ -0,0 +1,10 @@
+{
+  "run_id": "eval-5-without_skill",
+  "expectations": [
+    {"text": "Plan uses git worktree in a sibling directory", "passed": false, "evidence": "git checkout -b, no worktree"},
+    {"text": "References actual comment-checker hook files", "passed": true, "evidence": "Deep analysis of Go binary, tree-sitter, formatter.go, agent_memo.go with line numbers"},
+    {"text": "Adds test cases for Note: false positive scenarios", "passed": true, "evidence": "Detailed test cases distinguishing legit vs AI slop patterns"},
+    {"text": "Verification loop includes all 3 gates", "passed": false, "evidence": "Only bun test and typecheck. No review-work or Cubic."},
+    {"text": "Only modifies regex and adds tests — no unrelated changes", "passed": true, "evidence": "Adds allowed-prefix filter module — focused approach with config extension"}
+  ]
+}

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/without_skill/outputs/code-changes.md

@@ -0,0 +1,529 @@
+# Code Changes: comment-checker false positive fix
+
+## Change 1: Extend config schema
+
+**File: `src/config/schema/comment-checker.ts`**
+
+```typescript
+// BEFORE
+import { z } from "zod"
+
+export const CommentCheckerConfigSchema = z.object({
+  /** Custom prompt to replace the default warning message. Use {{comments}} placeholder for detected comments XML. */
+  custom_prompt: z.string().optional(),
+})
+
+export type CommentCheckerConfig = z.infer<typeof CommentCheckerConfigSchema>
+```
+
+```typescript
+// AFTER
+import { z } from "zod"
+
+const DEFAULT_ALLOWED_COMMENT_PREFIXES = [
+  "note:",
+  "todo:",
+  "fixme:",
+  "hack:",
+  "xxx:",
+  "warning:",
+  "important:",
+  "bug:",
+  "optimize:",
+  "workaround:",
+  "safety:",
+  "security:",
+  "perf:",
+  "see:",
+  "ref:",
+  "cf.",
+]
+
+export const CommentCheckerConfigSchema = z.object({
+  /** Custom prompt to replace the default warning message. Use {{comments}} placeholder for detected comments XML. */
+  custom_prompt: z.string().optional(),
+  /** Comment prefixes considered legitimate (not AI slop). Case-insensitive. Defaults include Note:, TODO:, FIXME:, etc. */
+  allowed_comment_prefixes: z.array(z.string()).optional().default(DEFAULT_ALLOWED_COMMENT_PREFIXES),
+})
+
+export type CommentCheckerConfig = z.infer<typeof CommentCheckerConfigSchema>
+```
+
+## Change 2: Create allowed-prefix-filter module
+
+**File: `src/hooks/comment-checker/allowed-prefix-filter.ts`** (NEW)
+
+```typescript
+const COMMENT_XML_REGEX = /<comment\s+line-number="\d+">([\s\S]*?)<\/comment>/g
+const COMMENTS_BLOCK_REGEX = /<comments\s+file="[^"]*">\s*([\s\S]*?)\s*<\/comments>/g
+const AGENT_MEMO_HEADER_REGEX = /🚨 AGENT MEMO COMMENT DETECTED.*?---\n\n/s
+
+function stripCommentPrefix(text: string): string {
+  let stripped = text.trim()
+  for (const prefix of ["//", "#", "/*", "--", "*"]) {
+    if (stripped.startsWith(prefix)) {
+      stripped = stripped.slice(prefix.length).trim()
+      break
+    }
+  }
+  return stripped
+}
+
+function isAllowedComment(commentText: string, allowedPrefixes: string[]): boolean {
+  const stripped = stripCommentPrefix(commentText).toLowerCase()
+  return allowedPrefixes.some((prefix) => stripped.startsWith(prefix.toLowerCase()))
+}
+
+function extractCommentTexts(xmlBlock: string): string[] {
+  const texts: string[] = []
+  let match: RegExpExecArray | null
+  const regex = new RegExp(COMMENT_XML_REGEX.source, COMMENT_XML_REGEX.flags)
+  while ((match = regex.exec(xmlBlock)) !== null) {
+    texts.push(match[1])
+  }
+  return texts
+}
+
+export function filterAllowedComments(
+  message: string,
+  allowedPrefixes: string[],
+): { hasRemainingComments: boolean; filteredMessage: string } {
+  if (!message || allowedPrefixes.length === 0) {
+    return { hasRemainingComments: true, filteredMessage: message }
+  }
+
+  const commentTexts = extractCommentTexts(message)
+
+  if (commentTexts.length === 0) {
+    return { hasRemainingComments: true, filteredMessage: message }
+  }
+
+  const disallowedComments = commentTexts.filter(
+    (text) => !isAllowedComment(text, allowedPrefixes),
+  )
+
+  if (disallowedComments.length === 0) {
+    return { hasRemainingComments: false, filteredMessage: "" }
+  }
+
+  if (disallowedComments.length === commentTexts.length) {
+    return { hasRemainingComments: true, filteredMessage: message }
+  }
+
+  let filteredMessage = message
+  for (const text of commentTexts) {
+    if (isAllowedComment(text, allowedPrefixes)) {
+      const escapedText = text.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")
+      const lineRegex = new RegExp(`\\s*<comment\\s+line-number="\\d+">${escapedText}</comment>\\n?`, "g")
+      filteredMessage = filteredMessage.replace(lineRegex, "")
+    }
+  }
+
+  filteredMessage = filteredMessage.replace(AGENT_MEMO_HEADER_REGEX, "")
+
+  return { hasRemainingComments: true, filteredMessage }
+}
+```
+
+## Change 3: Thread config through cli-runner.ts
+
+**File: `src/hooks/comment-checker/cli-runner.ts`**
+
+```typescript
+// BEFORE (processWithCli signature and body)
+export async function processWithCli(
+  input: { tool: string; sessionID: string; callID: string },
+  pendingCall: PendingCall,
+  output: { output: string },
+  cliPath: string,
+  customPrompt: string | undefined,
+  debugLog: (...args: unknown[]) => void,
+): Promise<void> {
+  await withCommentCheckerLock(async () => {
+    // ...
+    const result = await runCommentChecker(hookInput, cliPath, customPrompt)
+    if (result.hasComments && result.message) {
+      debugLog("CLI detected comments, appending message")
+      output.output += `\n\n${result.message}`
+    } else {
+      debugLog("CLI: no comments detected")
+    }
+  }, undefined, debugLog)
+}
+```
+
+```typescript
+// AFTER
+import { filterAllowedComments } from "./allowed-prefix-filter"
+
+export async function processWithCli(
+  input: { tool: string; sessionID: string; callID: string },
+  pendingCall: PendingCall,
+  output: { output: string },
+  cliPath: string,
+  customPrompt: string | undefined,
+  allowedPrefixes: string[],
+  debugLog: (...args: unknown[]) => void,
+): Promise<void> {
+  await withCommentCheckerLock(async () => {
+    void input
+    debugLog("using CLI mode with path:", cliPath)
+
+    const hookInput: HookInput = {
+      session_id: pendingCall.sessionID,
+      tool_name: pendingCall.tool.charAt(0).toUpperCase() + pendingCall.tool.slice(1),
+      transcript_path: "",
+      cwd: process.cwd(),
+      hook_event_name: "PostToolUse",
+      tool_input: {
+        file_path: pendingCall.filePath,
+        content: pendingCall.content,
+        old_string: pendingCall.oldString,
+        new_string: pendingCall.newString,
+        edits: pendingCall.edits,
+      },
+    }
+
+    const result = await runCommentChecker(hookInput, cliPath, customPrompt)
+
+    if (result.hasComments && result.message) {
+      const { hasRemainingComments, filteredMessage } = filterAllowedComments(
+        result.message,
+        allowedPrefixes,
+      )
+      if (hasRemainingComments && filteredMessage) {
+        debugLog("CLI detected comments, appending filtered message")
+        output.output += `\n\n${filteredMessage}`
+      } else {
+        debugLog("CLI: all detected comments matched allowed prefixes, suppressing")
+      }
+    } else {
+      debugLog("CLI: no comments detected")
+    }
+  }, undefined, debugLog)
+}
+
+// Same change applied to processApplyPatchEditsWithCli - add allowedPrefixes parameter
+export async function processApplyPatchEditsWithCli(
+  sessionID: string,
+  edits: ApplyPatchEdit[],
+  output: { output: string },
+  cliPath: string,
+  customPrompt: string | undefined,
+  allowedPrefixes: string[],
+  debugLog: (...args: unknown[]) => void,
+): Promise<void> {
+  debugLog("processing apply_patch edits:", edits.length)
+
+  for (const edit of edits) {
+    await withCommentCheckerLock(async () => {
+      const hookInput: HookInput = {
+        session_id: sessionID,
+        tool_name: "Edit",
+        transcript_path: "",
+        cwd: process.cwd(),
+        hook_event_name: "PostToolUse",
+        tool_input: {
+          file_path: edit.filePath,
+          old_string: edit.before,
+          new_string: edit.after,
+        },
+      }
+
+      const result = await runCommentChecker(hookInput, cliPath, customPrompt)
+
+      if (result.hasComments && result.message) {
+        const { hasRemainingComments, filteredMessage } = filterAllowedComments(
+          result.message,
+          allowedPrefixes,
+        )
+        if (hasRemainingComments && filteredMessage) {
+          debugLog("CLI detected comments for apply_patch file:", edit.filePath)
+          output.output += `\n\n${filteredMessage}`
+        }
+      }
+    }, undefined, debugLog)
+  }
+}
+```
+
+## Change 4: Update hook.ts to pass config
+
+**File: `src/hooks/comment-checker/hook.ts`**
+
+```typescript
+// BEFORE (in tool.execute.after handler, around line 177)
+await processWithCli(input, pendingCall, output, cliPath, config?.custom_prompt, debugLog)
+
+// AFTER
+const allowedPrefixes = config?.allowed_comment_prefixes ?? []
+await processWithCli(input, pendingCall, output, cliPath, config?.custom_prompt, allowedPrefixes, debugLog)
+```
+
+```typescript
+// BEFORE (in apply_patch section, around line 147-154)
+await processApplyPatchEditsWithCli(
+  input.sessionID,
+  edits,
+  output,
+  cliPath,
+  config?.custom_prompt,
+  debugLog,
+)
+
+// AFTER
+const allowedPrefixes = config?.allowed_comment_prefixes ?? []
+await processApplyPatchEditsWithCli(
+  input.sessionID,
+  edits,
+  output,
+  cliPath,
+  config?.custom_prompt,
+  allowedPrefixes,
+  debugLog,
+)
+```
+
+## Change 5: Test file for allowed-prefix-filter
+
+**File: `src/hooks/comment-checker/allowed-prefix-filter.test.ts`** (NEW)
+
+```typescript
+import { describe, test, expect } from "bun:test"
+
+import { filterAllowedComments } from "./allowed-prefix-filter"
+
+const DEFAULT_PREFIXES = [
+  "note:", "todo:", "fixme:", "hack:", "xxx:", "warning:",
+  "important:", "bug:", "optimize:", "workaround:", "safety:",
+  "security:", "perf:", "see:", "ref:", "cf.",
+]
+
+function buildMessage(comments: { line: number; text: string }[], filePath = "/tmp/test.ts"): string {
+  const xml = comments
+    .map((c) => `\t<comment line-number="${c.line}">${c.text}</comment>`)
+    .join("\n")
+  return `COMMENT/DOCSTRING DETECTED - IMMEDIATE ACTION REQUIRED\n\n` +
+    `Your recent changes contain comments or docstrings, which triggered this hook.\n` +
+    `Detected comments/docstrings:\n` +
+    `<comments file="${filePath}">\n${xml}\n</comments>\n`
+}
+
+describe("allowed-prefix-filter", () => {
+  describe("#given default allowed prefixes", () => {
+    describe("#when message contains only Note: comments", () => {
+      test("#then should suppress the entire message", () => {
+        const message = buildMessage([
+          { line: 5, text: "// Note: Thread-safe implementation" },
+          { line: 12, text: "// NOTE: See RFC 7231 for details" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(false)
+        expect(result.filteredMessage).toBe("")
+      })
+    })
+
+    describe("#when message contains only TODO/FIXME comments", () => {
+      test("#then should suppress the entire message", () => {
+        const message = buildMessage([
+          { line: 3, text: "// TODO: implement caching" },
+          { line: 7, text: "// FIXME: race condition here" },
+          { line: 15, text: "# HACK: workaround for upstream bug" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(false)
+        expect(result.filteredMessage).toBe("")
+      })
+    })
+
+    describe("#when message contains only AI slop comments", () => {
+      test("#then should keep the entire message", () => {
+        const message = buildMessage([
+          { line: 2, text: "// Added new validation logic" },
+          { line: 8, text: "// Refactored for better performance" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(true)
+        expect(result.filteredMessage).toBe(message)
+      })
+    })
+
+    describe("#when message contains mix of legitimate and slop comments", () => {
+      test("#then should keep message but remove allowed comment XML entries", () => {
+        const message = buildMessage([
+          { line: 5, text: "// Note: Thread-safe implementation" },
+          { line: 10, text: "// Changed from old API to new API" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(true)
+        expect(result.filteredMessage).not.toContain("Thread-safe implementation")
+        expect(result.filteredMessage).toContain("Changed from old API to new API")
+      })
+    })
+
+    describe("#when Note: comment has lowercase prefix", () => {
+      test("#then should still be treated as allowed (case-insensitive)", () => {
+        const message = buildMessage([
+          { line: 1, text: "// note: this is case insensitive" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(false)
+      })
+    })
+
+    describe("#when comment uses hash prefix", () => {
+      test("#then should strip prefix before matching", () => {
+        const message = buildMessage([
+          { line: 1, text: "# Note: Python style comment" },
+          { line: 5, text: "# TODO: something to do" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(false)
+      })
+    })
+
+    describe("#when comment has Security: prefix", () => {
+      test("#then should be treated as allowed", () => {
+        const message = buildMessage([
+          { line: 1, text: "// Security: validate input before processing" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(false)
+      })
+    })
+
+    describe("#when comment has Warning: prefix", () => {
+      test("#then should be treated as allowed", () => {
+        const message = buildMessage([
+          { line: 1, text: "// WARNING: This mutates the input array" },
+        ])
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(false)
+      })
+    })
+  })
+
+  describe("#given empty allowed prefixes", () => {
+    describe("#when any comments are detected", () => {
+      test("#then should pass through unfiltered", () => {
+        const message = buildMessage([
+          { line: 1, text: "// Note: this should pass through" },
+        ])
+
+        const result = filterAllowedComments(message, [])
+
+        expect(result.hasRemainingComments).toBe(true)
+        expect(result.filteredMessage).toBe(message)
+      })
+    })
+  })
+
+  describe("#given custom allowed prefixes", () => {
+    describe("#when comment matches custom prefix", () => {
+      test("#then should suppress it", () => {
+        const message = buildMessage([
+          { line: 1, text: "// PERF: O(n log n) complexity" },
+        ])
+
+        const result = filterAllowedComments(message, ["perf:"])
+
+        expect(result.hasRemainingComments).toBe(false)
+      })
+    })
+  })
+
+  describe("#given empty message", () => {
+    describe("#when filterAllowedComments is called", () => {
+      test("#then should return hasRemainingComments true with empty string", () => {
+        const result = filterAllowedComments("", DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(true)
+        expect(result.filteredMessage).toBe("")
+      })
+    })
+  })
+
+  describe("#given message with agent memo header", () => {
+    describe("#when all flagged comments are legitimate Note: comments", () => {
+      test("#then should suppress agent memo header along with comments", () => {
+        const message =
+          "🚨 AGENT MEMO COMMENT DETECTED - CODE SMELL ALERT 🚨\n\n" +
+          "⚠️  AGENT MEMO COMMENTS DETECTED - THIS IS A CODE SMELL  ⚠️\n\n" +
+          "You left \"memo-style\" comments...\n\n---\n\n" +
+          "Your recent changes contain comments...\n" +
+          "Detected comments/docstrings:\n" +
+          '<comments file="/tmp/test.ts">\n' +
+          '\t<comment line-number="5">// Note: Thread-safe</comment>\n' +
+          "</comments>\n"
+
+        const result = filterAllowedComments(message, DEFAULT_PREFIXES)
+
+        expect(result.hasRemainingComments).toBe(false)
+        expect(result.filteredMessage).toBe("")
+      })
+    })
+  })
+})
+```
+
+## Change 6: Update existing test for new parameter
+
+**File: `src/hooks/comment-checker/hook.apply-patch.test.ts`**
+
+The `processApplyPatchEditsWithCli` mock needs to account for the new `allowedPrefixes` parameter:
+
+```typescript
+// BEFORE (line 58)
+expect(processApplyPatchEditsWithCli).toHaveBeenCalledWith(
+  "ses_test",
+  [
+    { filePath: "/repo/src/a.ts", before: "const a = 1\n", after: "// comment\nconst a = 1\n" },
+    { filePath: "/repo/src/new.ts", before: "const b = 1\n", after: "// moved comment\nconst b = 1\n" },
+  ],
+  expect.any(Object),
+  "/tmp/fake-comment-checker",
+  undefined,
+  expect.any(Function),
+)
+
+// AFTER - add allowed_comment_prefixes argument
+expect(processApplyPatchEditsWithCli).toHaveBeenCalledWith(
+  "ses_test",
+  [
+    { filePath: "/repo/src/a.ts", before: "const a = 1\n", after: "// comment\nconst a = 1\n" },
+    { filePath: "/repo/src/new.ts", before: "const b = 1\n", after: "// moved comment\nconst b = 1\n" },
+  ],
+  expect.any(Object),
+  "/tmp/fake-comment-checker",
+  undefined,
+  expect.any(Array),
+  expect.any(Function),
+)
+```
+
+## Summary of all touched files
+
+| File | Action | Description |
+|------|--------|-------------|
+| `src/config/schema/comment-checker.ts` | Modified | Add `allowed_comment_prefixes` with defaults |
+| `src/hooks/comment-checker/allowed-prefix-filter.ts` | **New** | Post-processing filter for legitimate comment prefixes |
+| `src/hooks/comment-checker/allowed-prefix-filter.test.ts` | **New** | 11 test cases covering false positives and edge cases |
+| `src/hooks/comment-checker/cli-runner.ts` | Modified | Thread `allowedPrefixes` param, apply filter after binary result |
+| `src/hooks/comment-checker/hook.ts` | Modified | Pass `allowed_comment_prefixes` from config to CLI runner |
+| `src/hooks/comment-checker/hook.apply-patch.test.ts` | Modified | Update mock assertions for new parameter |

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/without_skill/outputs/execution-plan.md

@@ -0,0 +1,127 @@
+# Execution Plan: Relax comment-checker hook false positives
+
+## Problem Analysis
+
+The comment-checker hook delegates to an external Go binary (`code-yeongyu/go-claude-code-comment-checker`). The binary:
+1. Detects ALL comments in written/edited code using tree-sitter
+2. Filters out only BDD markers, linter directives, and shebangs
+3. Flags every remaining comment as problematic (exit code 2)
+4. In the output formatter (`formatter.go`), uses `AgentMemoFilter` to categorize comments for display
+
+The `AgentMemoFilter` in `pkg/filters/agent_memo.go` contains the overly aggressive regex:
+```go
+regexp.MustCompile(`(?i)^[\s#/*-]*note:\s*\w`),
+```
+
+This matches ANY comment starting with `Note:` (case-insensitive) followed by a word character, causing legitimate comments like `// Note: Thread-safe implementation` or `// NOTE: See RFC 7231` to be classified as "AGENT MEMO" AI slop with an aggressive warning banner.
+
+Additionally, the binary flags ALL non-filtered comments (not just agent memos), so even without the `Note:` regex, `// Note: ...` comments would still be flagged as generic "COMMENT DETECTED."
+
+## Architecture Understanding
+
+```
+TypeScript (oh-my-opencode)              Go Binary (go-claude-code-comment-checker)
+─────────────────────────────             ──────────────────────────────────────────
+hook.ts                                   main.go
+ ├─ tool.execute.before                    ├─ Read JSON from stdin
+ │   └─ registerPendingCall()              ├─ Detect comments (tree-sitter)
+ └─ tool.execute.after                     ├─ applyFilters (BDD, Directive, Shebang)
+     └─ processWithCli()                   ├─ FormatHookMessage (uses AgentMemoFilter for display)
+         └─ runCommentChecker()            └─ exit 0 (clean) or exit 2 (comments found, message on stderr)
+             └─ spawn binary, pipe JSON
+             └─ read stderr → message
+             └─ append to output
+```
+
+Key files in oh-my-opencode:
+- `src/hooks/comment-checker/hook.ts` - Hook factory, registers before/after handlers
+- `src/hooks/comment-checker/cli-runner.ts` - Orchestrates CLI invocation, semaphore
+- `src/hooks/comment-checker/cli.ts` - Binary resolution, process spawning, timeout handling
+- `src/hooks/comment-checker/types.ts` - PendingCall, CommentInfo types
+- `src/config/schema/comment-checker.ts` - Config schema (currently only `custom_prompt`)
+
+Key files in Go binary:
+- `pkg/filters/agent_memo.go` - Contains the aggressive `note:\s*\w` regex (line 20)
+- `pkg/output/formatter.go` - Uses AgentMemoFilter to add "AGENT MEMO" warnings
+- `cmd/comment-checker/main.go` - Filter pipeline (BDD + Directive + Shebang only)
+
+## Step-by-Step Plan
+
+### Step 1: Create feature branch
+```bash
+git checkout dev
+git pull origin dev
+git checkout -b fix/comment-checker-note-false-positive
+```
+
+### Step 2: Extend CommentCheckerConfigSchema
+**File: `src/config/schema/comment-checker.ts`**
+
+Add `allowed_comment_prefixes` field with sensible defaults. This lets users configure which comment prefixes should be treated as legitimate (not AI slop).
+
+### Step 3: Add a post-processing filter in cli-runner.ts
+**File: `src/hooks/comment-checker/cli-runner.ts`**
+
+After the Go binary returns its result, parse the stderr message to identify and suppress comments that match allowed prefixes. The binary's output contains XML like:
+```xml
+<comments file="/path/to/file.ts">
+  <comment line-number="5">// Note: Thread-safe</comment>
+</comments>
+```
+
+Add a function `filterAllowedComments()` that:
+1. Extracts `<comment>` elements from the message
+2. Checks if the comment text matches any allowed prefix pattern
+3. If ALL flagged comments match allowed patterns, suppress the entire warning
+4. If some comments are legitimate and some aren't, rebuild the message without the legitimate ones
+
+### Step 4: Create dedicated filter module
+**File: `src/hooks/comment-checker/allowed-prefix-filter.ts`** (new)
+
+Extract the filtering logic into its own module per the 200 LOC / single-responsibility rule.
+
+### Step 5: Pass allowed_comment_prefixes through the hook chain
+**File: `src/hooks/comment-checker/hook.ts`**
+
+Thread the `allowed_comment_prefixes` config from `createCommentCheckerHooks()` down to `processWithCli()` and `processApplyPatchEditsWithCli()`.
+
+### Step 6: Add test cases
+**File: `src/hooks/comment-checker/allowed-prefix-filter.test.ts`** (new)
+
+Test cases covering:
+- `// Note: Thread-safe implementation` - should NOT be flagged (false positive)
+- `// NOTE: See RFC 7231 for details` - should NOT be flagged
+- `// Note: changed from X to Y` - SHOULD still be flagged (genuine AI slop)
+- `// TODO: implement caching` - should NOT be flagged
+- `// FIXME: race condition` - should NOT be flagged
+- `// HACK: workaround for upstream bug` - should NOT be flagged
+- `// Added new validation logic` - SHOULD be flagged
+- Custom allowed patterns from config
+
+**File: `src/hooks/comment-checker/cli-runner.test.ts`** (new or extend cli.test.ts)
+
+Integration-level tests for the post-processing pipeline.
+
+### Step 7: Verify
+```bash
+bun test src/hooks/comment-checker/
+bun run typecheck
+```
+
+### Step 8: Commit and push
+```bash
+git add -A
+git commit -m "fix(comment-checker): add allowed-prefix filter to reduce false positives on Note: comments"
+git push -u origin fix/comment-checker-note-false-positive
+```
+
+### Step 9: Create PR
+```bash
+gh pr create --title "fix(comment-checker): reduce false positives for legitimate Note: comments" --body-file /tmp/pr-body.md --base dev
+```
+
+### Step 10 (Follow-up): Upstream Go binary fix
+File an issue or PR on `code-yeongyu/go-claude-code-comment-checker` to:
+1. Relax `(?i)^[\s#/*-]*note:\s*\w` to be more specific (e.g., `note:\s*(changed|modified|updated|added|removed|implemented|refactored)`)
+2. Add a dedicated `LegitimateCommentFilter` to the filter pipeline in `main.go`
+3. Support `--allow-prefix` CLI flag for external configuration

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/without_skill/outputs/pr-description.md

@@ -0,0 +1,42 @@
+## Summary
+
+- Add `allowed_comment_prefixes` config to `CommentCheckerConfigSchema` with sensible defaults (Note:, TODO:, FIXME:, HACK:, WARNING:, etc.)
+- Add post-processing filter in `allowed-prefix-filter.ts` that suppresses false positives from the Go binary's output before appending to tool output
+- Add 11 test cases covering false positive scenarios (Note:, TODO:, FIXME:, case-insensitivity, mixed comments, agent memo header suppression)
+
+## Problem
+
+The comment-checker hook's upstream Go binary (`go-claude-code-comment-checker`) flags ALL non-filtered comments as problematic. Its `AgentMemoFilter` regex `(?i)^[\s#/*-]*note:\s*\w` classifies any `Note:` comment as AI-generated "agent memo" slop, triggering an aggressive warning banner.
+
+This causes false positives for legitimate, widely-used comment patterns:
+```typescript
+// Note: Thread-safe implementation required due to concurrent access
+// NOTE: See RFC 7231 section 6.5.4 for 404 semantics
+// Note: This timeout matches the upstream service SLA
+```
+
+These are standard engineering documentation patterns, not AI slop.
+
+## Solution
+
+Rather than waiting for an upstream binary fix, this PR adds a configurable **post-processing filter** on the TypeScript side:
+
+1. **Config**: `comment_checker.allowed_comment_prefixes` - array of case-insensitive prefixes (defaults: `note:`, `todo:`, `fixme:`, `hack:`, `warning:`, `important:`, `bug:`, etc.)
+2. **Filter**: After the Go binary returns flagged comments, `filterAllowedComments()` parses the XML output and suppresses comments matching allowed prefixes
+3. **Behavior**: If ALL flagged comments are legitimate → suppress entire warning. If mixed → remove only the legitimate entries from the XML, keep the warning for actual slop.
+
+Users can customize via config:
+```jsonc
+{
+  "comment_checker": {
+    "allowed_comment_prefixes": ["note:", "todo:", "fixme:", "custom-prefix:"]
+  }
+}
+```
+
+## Test Plan
+
+- 11 new test cases in `allowed-prefix-filter.test.ts`
+- Updated assertion in `hook.apply-patch.test.ts` for new parameter
+- `bun test src/hooks/comment-checker/` passes
+- `bun run typecheck` clean

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/without_skill/outputs/verification-strategy.md

@@ -0,0 +1,120 @@
+# Verification Strategy
+
+## 1. Unit Tests
+
+### New test file: `allowed-prefix-filter.test.ts`
+Run: `bun test src/hooks/comment-checker/allowed-prefix-filter.test.ts`
+
+| # | Scenario | Input | Expected |
+|---|----------|-------|----------|
+| 1 | Only Note: comments (default prefixes) | `// Note: Thread-safe`, `// NOTE: See RFC` | `hasRemainingComments: false`, empty message |
+| 2 | Only TODO/FIXME/HACK (default prefixes) | `// TODO: impl`, `// FIXME: race`, `# HACK: workaround` | Suppressed |
+| 3 | Only AI slop comments | `// Added validation`, `// Refactored for perf` | Full message preserved |
+| 4 | Mixed legitimate + slop | `// Note: Thread-safe`, `// Changed from old to new` | Message kept, Note: entry removed from XML |
+| 5 | Case-insensitive Note: | `// note: lowercase test` | Suppressed |
+| 6 | Hash-prefixed comments | `# Note: Python`, `# TODO: something` | Suppressed (prefix stripped before matching) |
+| 7 | Security: prefix | `// Security: validate input` | Suppressed |
+| 8 | Warning: prefix | `// WARNING: mutates input` | Suppressed |
+| 9 | Empty allowed prefixes | `// Note: should pass through` | Full message preserved (no filtering) |
+| 10 | Custom prefix | `// PERF: O(n log n)` with `["perf:"]` | Suppressed |
+| 11 | Agent memo header + Note: | Full agent memo banner + `// Note: Thread-safe` | Entire message suppressed including banner |
+
+### Existing test: `hook.apply-patch.test.ts`
+Run: `bun test src/hooks/comment-checker/hook.apply-patch.test.ts`
+
+Verify the updated mock assertion accepts the new `allowedPrefixes` array parameter.
+
+### Existing test: `cli.test.ts`
+Run: `bun test src/hooks/comment-checker/cli.test.ts`
+
+Verify no regressions in binary spawning, timeout, and semaphore logic.
+
+## 2. Type Checking
+
+```bash
+bun run typecheck
+```
+
+Verify:
+- `CommentCheckerConfigSchema` change propagates correctly to `CommentCheckerConfig` type
+- All call sites in `hook.ts` and `cli-runner.ts` pass the new parameter
+- `filterAllowedComments` return type matches usage in `cli-runner.ts`
+- No new type errors introduced
+
+## 3. LSP Diagnostics
+
+```bash
+# Check all changed files for errors
+lsp_diagnostics src/config/schema/comment-checker.ts
+lsp_diagnostics src/hooks/comment-checker/allowed-prefix-filter.ts
+lsp_diagnostics src/hooks/comment-checker/cli-runner.ts
+lsp_diagnostics src/hooks/comment-checker/hook.ts
+lsp_diagnostics src/hooks/comment-checker/allowed-prefix-filter.test.ts
+```
+
+## 4. Full Test Suite
+
+```bash
+bun test src/hooks/comment-checker/
+```
+
+All 4 test files should pass:
+- `cli.test.ts` (existing - no regressions)
+- `pending-calls.test.ts` (existing - no regressions)
+- `hook.apply-patch.test.ts` (modified assertion)
+- `allowed-prefix-filter.test.ts` (new - all 11 cases)
+
+## 5. Build Verification
+
+```bash
+bun run build
+```
+
+Ensure the new module is properly bundled and exported.
+
+## 6. Integration Verification (Manual)
+
+If binary is available locally:
+
+```bash
+# Test with a file containing Note: comment
+echo '{"session_id":"test","tool_name":"Write","transcript_path":"","cwd":"/tmp","hook_event_name":"PostToolUse","tool_input":{"file_path":"/tmp/test.ts","content":"// Note: Thread-safe implementation\nconst x = 1"}}' | ~/.cache/oh-my-opencode/bin/comment-checker check
+echo "Exit code: $?"
+```
+
+Expected: Binary returns exit 2 (comment detected), but the TypeScript post-filter should suppress it.
+
+## 7. Config Validation
+
+Test that config changes work:
+
+```jsonc
+// .opencode/oh-my-opencode.jsonc
+{
+  "comment_checker": {
+    // Override: only allow Note: and TODO:
+    "allowed_comment_prefixes": ["note:", "todo:"]
+  }
+}
+```
+
+Verify Zod schema accepts the config and defaults are applied when field is omitted.
+
+## 8. Regression Checks
+
+Verify the following still work correctly:
+- AI slop comments (`// Added new feature`, `// Refactored for performance`) are still flagged
+- BDD comments (`// given`, `// when`, `// then`) are still allowed (binary-side filter)
+- Linter directives (`// eslint-disable`, `// @ts-ignore`) are still allowed (binary-side filter)
+- Shebangs (`#!/usr/bin/env node`) are still allowed (binary-side filter)
+- `custom_prompt` config still works
+- Semaphore prevents concurrent comment-checker runs
+- Timeout handling (30s) still works
+
+## 9. Edge Cases to Watch
+
+- Empty message from binary (exit code 0) - filter should be no-op
+- Binary not available - hook gracefully degrades (existing behavior)
+- Message with no `<comment>` XML elements - filter passes through
+- Very long messages with many comments - regex performance
+- Comments containing XML-special characters (`<`, `>`, `&`) in text

.opencode/skills/work-with-pr-workspace/iteration-1/eval-5/without_skill/timing.json

@@ -0,0 +1 @@
+{"total_tokens": null, "duration_ms": 399000, "total_duration_seconds": 399}

.opencode/skills/work-with-pr/SKILL.md

@@ -0,0 +1,348 @@
+---
+name: work-with-pr
+description: "Full PR lifecycle: git worktree → implement → atomic commits → PR creation → verification loop (CI + review-work + Cubic approval) → merge. Keeps iterating until ALL gates pass and PR is merged. Worktree auto-cleanup after merge. Use whenever implementation work needs to land as a PR. Triggers: 'create a PR', 'implement and PR', 'work on this and make a PR', 'implement issue', 'land this as a PR', 'work-with-pr', 'PR workflow', 'implement end to end', even when user just says 'implement X' if the context implies PR delivery."
+---
+
+# Work With PR — Full PR Lifecycle
+
+You are executing a complete PR lifecycle: from isolated worktree setup through implementation, PR creation, and an unbounded verification loop until the PR is merged. The loop has three gates — CI, review-work, and Cubic — and you keep fixing and pushing until all three pass simultaneously.
+
+<architecture>
+
+```
+Phase 0: Setup         → Branch + worktree in sibling directory
+Phase 1: Implement     → Do the work, atomic commits
+Phase 2: PR Creation   → Push, create PR targeting dev
+Phase 3: Verify Loop   → Unbounded iteration until ALL gates pass:
+  ├─ Gate A: CI         → gh pr checks (bun test, typecheck, build)
+  ├─ Gate B: review-work → 5-agent parallel review
+  └─ Gate C: Cubic      → cubic-dev-ai[bot] "No issues found"
+Phase 4: Merge         → Squash merge, worktree cleanup
+```
+
+</architecture>
+
+---
+
+## Phase 0: Setup
+
+Create an isolated worktree so the user's main working directory stays clean. This matters because the user may have uncommitted work, and checking out a branch would destroy it.
+
+<setup>
+
+### 1. Resolve repository context
+
+```bash
+REPO=$(gh repo view --json nameWithOwner -q .nameWithOwner)
+REPO_NAME=$(basename "$PWD")
+BASE_BRANCH="dev"  # CI blocks PRs to master
+```
+
+### 2. Create branch
+
+If user provides a branch name, use it. Otherwise, derive from the task:
+
+```bash
+# Auto-generate: feature/short-description or fix/short-description
+BRANCH_NAME="feature/$(echo "$TASK_SUMMARY" | tr '[:upper:] ' '[:lower:]-' | head -c 50)"
+git fetch origin "$BASE_BRANCH"
+git branch "$BRANCH_NAME" "origin/$BASE_BRANCH"
+```
+
+### 3. Create worktree
+
+Place worktrees as siblings to the repo — not inside it. This avoids git nested repo issues and keeps the working tree clean.
+
+```bash
+WORKTREE_PATH="../${REPO_NAME}-wt/${BRANCH_NAME}"
+mkdir -p "$(dirname "$WORKTREE_PATH")"
+git worktree add "$WORKTREE_PATH" "$BRANCH_NAME"
+```
+
+### 4. Set working context
+
+All subsequent work happens inside the worktree. Install dependencies if needed:
+
+```bash
+cd "$WORKTREE_PATH"
+# If bun project:
+[ -f "bun.lock" ] && bun install
+```
+
+</setup>
+
+---
+
+## Phase 1: Implement
+
+Do the actual implementation work inside the worktree. The agent using this skill does the work directly — no subagent delegation for the implementation itself.
+
+**Scope discipline**: For bug fixes, stay minimal. Fix the bug, add a test for it, done. Do not refactor surrounding code, add config options, or "improve" things that aren't broken. The verification loop will catch regressions — trust the process.
+
+<implementation>
+
+### Commit strategy
+
+Use the git-master skill's atomic commit principles. The reason for atomic commits: if CI fails on one change, you can isolate and fix it without unwinding everything.
+
+```
+3+ files changed  → 2+ commits minimum
+5+ files changed  → 3+ commits minimum
+10+ files changed → 5+ commits minimum
+```
+
+Each commit should pair implementation with its tests. Load `git-master` skill when committing:
+
+```
+task(category="quick", load_skills=["git-master"], prompt="Commit the changes atomically following git-master conventions. Repository is at {WORKTREE_PATH}.")
+```
+
+### Pre-push local validation
+
+Before pushing, run the same checks CI will run. Catching failures locally saves a full CI round-trip (~3-5 min):
+
+```bash
+bun run typecheck
+bun test
+bun run build
+```
+
+Fix any failures before pushing. Each fix-commit cycle should be atomic.
+
+</implementation>
+
+---
+
+## Phase 2: PR Creation
+
+<pr_creation>
+
+### Push and create PR
+
+```bash
+git push -u origin "$BRANCH_NAME"
+```
+
+Create the PR using the project's template structure:
+
+```bash
+gh pr create \
+  --base "$BASE_BRANCH" \
+  --head "$BRANCH_NAME" \
+  --title "$PR_TITLE" \
+  --body "$(cat <<'EOF'
+## Summary
+[1-3 sentences describing what this PR does and why]
+
+## Changes
+[Bullet list of key changes]
+
+## Testing
+- `bun run typecheck` ✅
+- `bun test` ✅
+- `bun run build` ✅
+
+## Related Issues
+[Link to issue if applicable]
+EOF
+)"
+```
+
+Capture the PR number:
+
+```bash
+PR_NUMBER=$(gh pr view --json number -q .number)
+```
+
+</pr_creation>
+
+---
+
+## Phase 3: Verification Loop
+
+This is the core of the skill. Three gates must ALL pass for the PR to be ready. The loop has no iteration cap — keep going until done. Gate ordering is intentional: CI is cheapest/fastest, review-work is most thorough, Cubic is external and asynchronous.
+
+<verify_loop>
+
+```
+while true:
+  1. Wait for CI          → Gate A
+  2. If CI fails          → read logs, fix, commit, push, continue
+  3. Run review-work      → Gate B
+  4. If review fails      → fix blocking issues, commit, push, continue
+  5. Check Cubic          → Gate C
+  6. If Cubic has issues   → fix issues, commit, push, continue
+  7. All three pass       → break
+```
+
+### Gate A: CI Checks
+
+CI is the fastest feedback loop. Wait for it to complete, then parse results.
+
+```bash
+# Wait for checks to start (GitHub needs a moment after push)
+# Then watch for completion
+gh pr checks "$PR_NUMBER" --watch --fail-fast
+```
+
+**On failure**: Get the failed run logs to understand what broke:
+
+```bash
+# Find the failed run
+RUN_ID=$(gh run list --branch "$BRANCH_NAME" --status failure --json databaseId --jq '.[0].databaseId')
+
+# Get failed job logs
+gh run view "$RUN_ID" --log-failed
+```
+
+Read the logs, fix the issue, commit atomically, push, and re-enter the loop.
+
+### Gate B: review-work
+
+The review-work skill launches 5 parallel sub-agents (goal verification, QA, code quality, security, context mining). All 5 must pass.
+
+Invoke review-work after CI passes — there's no point reviewing code that doesn't build:
+
+```
+task(
+  category="unspecified-high",
+  load_skills=["review-work"],
+  run_in_background=false,
+  description="Post-implementation review of PR changes",
+  prompt="Review the implementation work on branch {BRANCH_NAME}. The worktree is at {WORKTREE_PATH}. Goal: {ORIGINAL_GOAL}. Constraints: {CONSTRAINTS}. Run command: bun run dev (or as appropriate)."
+)
+```
+
+**On failure**: review-work reports blocking issues with specific files and line numbers. Fix each blocking issue, commit, push, and re-enter the loop from Gate A (since code changed, CI must re-run).
+
+### Gate C: Cubic Approval
+
+Cubic (`cubic-dev-ai[bot]`) is an automated review bot that comments on PRs. It does NOT use GitHub's APPROVED review state — instead it posts comments with issue counts and confidence scores.
+
+**Approval signal**: The latest Cubic comment contains `**No issues found**` and confidence `**5/5**`.
+
+**Issue signal**: The comment lists issues with file-level detail.
+
+```bash
+# Get the latest Cubic review
+CUBIC_REVIEW=$(gh api "repos/${REPO}/pulls/${PR_NUMBER}/reviews" \
+  --jq '[.[] | select(.user.login == "cubic-dev-ai[bot]")] | last | .body')
+
+# Check if approved
+if echo "$CUBIC_REVIEW" | grep -q "No issues found"; then
+  echo "Cubic: APPROVED"
+else
+  echo "Cubic: ISSUES FOUND"
+  echo "$CUBIC_REVIEW"
+fi
+```
+
+**On issues**: Cubic's review body contains structured issue descriptions. Parse them, determine which are valid (some may be false positives), fix the valid ones, commit, push, re-enter from Gate A.
+
+Cubic reviews are triggered automatically on PR updates. After pushing a fix, wait for the new review to appear before checking again. Use `gh api` polling with a conditional loop:
+
+```bash
+# Wait for new Cubic review after push
+PUSH_TIME=$(date -u +%Y-%m-%dT%H:%M:%SZ)
+while true; do
+  LATEST_REVIEW_TIME=$(gh api "repos/${REPO}/pulls/${PR_NUMBER}/reviews" \
+    --jq '[.[] | select(.user.login == "cubic-dev-ai[bot]")] | last | .submitted_at')
+  if [[ "$LATEST_REVIEW_TIME" > "$PUSH_TIME" ]]; then
+    break
+  fi
+  # Use gh api call itself as the delay mechanism — each call takes ~1-2s
+  # For longer waits, use: timeout 30 gh pr checks "$PR_NUMBER" --watch 2>/dev/null || true
+done
+```
+
+### Iteration discipline
+
+Each iteration through the loop:
+1. Fix ONLY the issues identified by the failing gate
+2. Commit atomically (one logical fix per commit)
+3. Push
+4. Re-enter from Gate A (code changed → full re-verification)
+
+Avoid the temptation to "improve" unrelated code during fix iterations. Scope creep in the fix loop makes debugging harder and can introduce new failures.
+
+</verify_loop>
+
+---
+
+## Phase 4: Merge & Cleanup
+
+Once all three gates pass:
+
+<merge_cleanup>
+
+### Merge the PR
+
+```bash
+# Squash merge to keep history clean
+gh pr merge "$PR_NUMBER" --squash --delete-branch
+```
+
+### Clean up the worktree
+
+The worktree served its purpose — remove it to avoid disk bloat:
+
+```bash
+cd "$ORIGINAL_DIR"  # Return to original working directory
+git worktree remove "$WORKTREE_PATH"
+# Prune any stale worktree references
+git worktree prune
+```
+
+### Report completion
+
+Summarize what happened:
+
+```
+## PR Merged ✅
+
+- **PR**: #{PR_NUMBER} — {PR_TITLE}
+- **Branch**: {BRANCH_NAME} → {BASE_BRANCH}
+- **Iterations**: {N} verification loops
+- **Gates passed**: CI ✅ | review-work ✅ | Cubic ✅
+- **Worktree**: cleaned up
+```
+
+</merge_cleanup>
+
+---
+
+## Failure Recovery
+
+<failure_recovery>
+
+If you hit an unrecoverable error (e.g., merge conflict with base branch, infrastructure failure):
+
+1. **Do NOT delete the worktree** — the user may want to inspect or continue manually
+2. Report what happened, what was attempted, and where things stand
+3. Include the worktree path so the user can resume
+
+For merge conflicts:
+
+```bash
+cd "$WORKTREE_PATH"
+git fetch origin "$BASE_BRANCH"
+git rebase "origin/$BASE_BRANCH"
+# Resolve conflicts, then continue the loop
+```
+
+</failure_recovery>
+
+---
+
+## Anti-Patterns
+
+| Violation | Why it fails | Severity |
+|-----------|-------------|----------|
+| Working in main worktree instead of isolated worktree | Pollutes user's working directory, may destroy uncommitted work | CRITICAL |
+| Pushing directly to dev/master | Bypasses review entirely | CRITICAL |
+| Skipping CI gate after code changes | review-work and Cubic may pass on stale code | CRITICAL |
+| Fixing unrelated code during verification loop | Scope creep causes new failures | HIGH |
+| Deleting worktree on failure | User loses ability to inspect/resume | HIGH |
+| Ignoring Cubic false positives without justification | Cubic issues should be evaluated, not blindly dismissed | MEDIUM |
+| Giant single commits | Harder to isolate failures, violates git-master principles | MEDIUM |
+| Not running local checks before push | Wastes CI time on obvious failures | MEDIUM |

AGENTS.md

@@ -4,7 +4,7 @@
 
 ## OVERVIEW
 
-OpenCode plugin (npm: `oh-my-opencode`) that extends Claude Code (OpenCode fork) with multi-agent orchestration, 46 lifecycle hooks, 26 tools, skill/command/MCP systems, and Claude Code compatibility. 1268 TypeScript files, 160k LOC.
+OpenCode plugin (npm: `oh-my-opencode`) that extends Claude Code (OpenCode fork) with multi-agent orchestration, 48 lifecycle hooks, 26 tools, skill/command/MCP systems, and Claude Code compatibility. 1268 TypeScript files, 160k LOC.
 
 ## STRUCTURE
 
@@ -14,14 +14,14 @@ oh-my-opencode/
 │   ├── index.ts              # Plugin entry: loadConfig → createManagers → createTools → createHooks → createPluginInterface
 │   ├── plugin-config.ts      # JSONC multi-level config: user → project → defaults (Zod v4)
 │   ├── agents/               # 11 agents (Sisyphus, Hephaestus, Oracle, Librarian, Explore, Atlas, Prometheus, Metis, Momus, Multimodal-Looker, Sisyphus-Junior)
-│   ├── hooks/                # 46 hooks across 45 directories + 11 standalone files
+│   ├── hooks/                # 48 lifecycle hooks across dedicated modules and standalone files
 │   ├── tools/                # 26 tools across 15 directories
 │   ├── features/             # 19 feature modules (background-agent, skill-loader, tmux, MCP-OAuth, etc.)
 │   ├── shared/               # 95+ utility files in 13 categories
 │   ├── config/               # Zod v4 schema system (24 files)
 │   ├── cli/                  # CLI: install, run, doctor, mcp-oauth (Commander.js)
 │   ├── mcp/                  # 3 built-in remote MCPs (websearch, context7, grep_app)
-│   ├── plugin/               # 8 OpenCode hook handlers + 46 hook composition
+│   ├── plugin/               # 8 OpenCode hook handlers + 48 hook composition
 │   └── plugin-handlers/      # 6-phase config loading pipeline
 ├── packages/                 # Monorepo: cli-runner, 12 platform binaries
 └── local-ignore/             # Dev-only test fixtures
@@ -34,7 +34,7 @@ OhMyOpenCodePlugin(ctx)
   ├─→ loadPluginConfig()         # JSONC parse → project/user merge → Zod validate → migrate
   ├─→ createManagers()           # TmuxSessionManager, BackgroundManager, SkillMcpManager, ConfigHandler
   ├─→ createTools()              # SkillContext + AvailableCategories + ToolRegistry (26 tools)
-  ├─→ createHooks()              # 3-tier: Core(37) + Continuation(7) + Skill(2) = 46 hooks
+  ├─→ createHooks()              # 3-tier: Core(39) + Continuation(7) + Skill(2) = 48 hooks
   └─→ createPluginInterface()    # 8 OpenCode hook handlers → PluginInterface
 ```
 
@@ -97,7 +97,7 @@ Fields: agents (14 overridable, 21 fields each), categories (8 built-in + custom
 - **Test pattern**: Bun test (`bun:test`), co-located `*.test.ts`, given/when/then style (nested describe with `#given`/`#when`/`#then` prefixes)
 - **CI test split**: mock-heavy tests run in isolation (separate `bun test` processes), rest in batch
 - **Factory pattern**: `createXXX()` for all tools, hooks, agents
-- **Hook tiers**: Session (23) → Tool-Guard (10) → Transform (4) → Continuation (7) → Skill (2)
+- **Hook tiers**: Session (23) → Tool-Guard (12) → Transform (4) → Continuation (7) → Skill (2)
 - **Agent modes**: `primary` (respects UI model) vs `subagent` (own fallback chain) vs `all`
 - **Model resolution**: 4-step: override → category-default → provider-fallback → system-default
 - **Config format**: JSONC with comments, Zod v4 validation, snake_case keys

FIX-BLOCKS.md

@@ -0,0 +1,122 @@
+# Pre-Publish BLOCK Issues: Fix ALL Before Release
+
+Two independent pre-publish reviews (Opus 4.6 + GPT-5.4) both concluded **BLOCK -- do not publish**. You must fix ALL blocking issues below using UltraBrain parallel agents. Work TDD-style: write/update tests first, then fix, verify tests pass.
+
+## Strategy
+
+Use ultrawork (ulw) to spawn UltraBrain agents in parallel. Each UB agent gets a non-overlapping scope. After all agents complete, run bun test to verify everything passes. Commit atomically per fix group.
+
+---
+
+## CRITICAL BLOCKERS (must fix -- 6 items)
+
+### C1: Hashline Backward Compatibility
+**Problem:** Strict whitespace hashing in hashline changes LINE#ID values for indented lines. Breaks existing anchors in cached/persisted edit operations.
+**Fix:** Add a compatibility shim -- when lookup by new hash fails, fall back to legacy hash (without strict whitespace). Or version the hash format.
+**Files:** Look for hashline-related files in src/tools/ or src/shared/
+
+### C2: OpenAI-Only Model Catalog Broken with OpenCode-Go
+**Problem:** isOpenAiOnlyAvailability() does not exclude availability.opencodeGo. When OpenCode-Go is present, OpenAI-only detection is wrong -- models get misrouted.
+**Fix:** Add !availability.opencodeGo check to isOpenAiOnlyAvailability().
+**Files:** Model/provider system files -- search for isOpenAiOnlyAvailability
+
+### C3: CLI/Runtime Model Table Divergence
+**Problem:** Model tables disagree between CLI install-time and runtime:
+- ultrabrain: gpt-5.3-codex in CLI vs gpt-5.4 in runtime
+- atlas: claude-sonnet-4-5 in CLI vs claude-sonnet-4-6 in runtime
+- unspecified-high also diverges
+**Fix:** Reconcile all model tables. Pick the correct model for each and make CLI + runtime match.
+**Files:** Search for model table definitions, agent configs, CLI model references
+
+### C4: atlas/metis/sisyphus-junior Missing OpenAI Fallbacks
+**Problem:** These agents can resolve to opencode/glm-4.7-free or undefined in OpenAI-only environments. No valid OpenAI fallback paths exist.
+**Fix:** Add valid OpenAI model fallback paths for all agents that need them.
+**Files:** Agent config/model resolution code
+
+### C5: model_fallback Default Mismatch
+**Problem:** Schema and docs say model_fallback defaults to false, but runtime treats unset as true. Silent behavior change for all users.
+**Fix:** Align -- either update schema/docs to say true, or fix runtime to default to false. Check what the intended behavior is from git history.
+**Files:** Schema definition, runtime config loading
+
+### C6: background_output Default Changed
+**Problem:** background_output now defaults to full_session=true. Old callers get different output format without code changes.
+**Fix:** Either document this change clearly, or restore old default and make full_session opt-in.
+**Files:** Background output handling code
+
+---
+
+## HIGH PRIORITY (strongly recommended -- 4 items)
+
+### H1: Runtime Fallback session-status-handler Race
+**Problem:** When fallback model is already pending, the handler cannot advance the chain on subsequent cooldown events.
+**Fix:** Allow override like message-update-handler does.
+**Files:** Search for session-status-handler, message-update-handler
+
+### H2: Atlas Final-Wave Approval Gate Logic
+**Problem:** Approval gate logic does not match real Prometheus plan structure (nested checkboxes, parallel execution). Trigger logic is wrong.
+**Fix:** Update to handle real plan structures.
+**Files:** Atlas agent code, approval gate logic
+
+### H3: delegate-task-english-directive Dead Code
+**Problem:** Not dispatched from tool-execute-before.ts + wrong hook signature. Either wire properly or remove entirely.
+**Fix:** Remove if not needed (cleaner). If needed, fix dispatch + signature.
+**Files:** src/hooks/, tool-execute-before.ts
+
+### H4: Auto-Slash-Command Session-Lifetime Dedup
+**Problem:** Dedup uses session lifetime, suppressing legitimate repeated identical commands.
+**Fix:** Change to short TTL (e.g., 30 seconds) instead of session lifetime.
+**Files:** Slash command handling code
+
+---
+
+## ADDITIONAL BLOCKERS FROM GPT-5.4 REVIEW
+
+### G1: Package Identity Split-Brain
+**Problem:** Installer writes oh-my-openagent but doctor, auto-update, version lookup, publish workflow still reference oh-my-opencode. Half-migrated state.
+**Fix:** Audit ALL references to package name. Either complete the migration consistently or revert to single name for this release.
+**Files:** Installer, doctor, auto-update, version lookup, publish workflow -- grep for both package names
+
+### G2: OpenCode-Go --opencode-go Value Validation
+**Problem:** No validation for --opencode-go CLI value. No detection of existing OpenCode-Go installations.
+**Fix:** Add value validation + existing install detection.
+**Files:** CLI option handling code
+
+### G3: Skill/Hook Reference Errors
+**Problem:**
+- work-with-pr references non-existent git tool category
+- github-triage references TaskCreate/TaskUpdate which are not real tool names
+**Fix:** Fix tool references to use actual tool names.
+**Files:** Skill definition files in .opencode/skills/
+
+### G4: Stale Context-Limit Cache
+**Problem:** Shared context-limit resolver caches provider config. When config changes, stale removed limits persist and corrupt compaction/truncation decisions.
+**Fix:** Add cache invalidation when provider config changes, or make the resolver stateless.
+**Files:** Context-limit resolver, compaction code
+
+### G5: disabled_hooks Schema vs Runtime Contract Mismatch
+**Problem:** Schema is strict (rejects unknown hook names) but runtime is permissive (ignores unknown). Contract disagreement.
+**Fix:** Align -- either make both strict or both permissive.
+**Files:** Hook schema definition, runtime hook loading
+
+---
+
+## EXECUTION INSTRUCTIONS
+
+1. Spawn UltraBrain agents to fix these in parallel -- group by file proximity:
+   - UB-1: C1 (hashline) + H4 (slash-command dedup)
+   - UB-2: C2 + C3 + C4 (model/provider system) + G2
+   - UB-3: C5 + C6 (config defaults) + G5
+   - UB-4: H1 + H2 (runtime handlers + Atlas gate)
+   - UB-5: H3 + G3 (dead code + skill references)
+   - UB-6: G1 (package identity -- full audit)
+   - UB-7: G4 (context-limit cache)
+
+2. Each UB agent MUST:
+   - Write or update tests FIRST (TDD)
+   - Implement the fix
+   - Run bun test on affected test files
+   - Commit with descriptive message
+
+3. After all UB agents complete, run full bun test to verify no regressions.
+
+ulw

README.ja.md

@@ -4,6 +4,17 @@
 > コアメンテナーのQが負傷したため、今週は Issue/PR への返信とリリースが遅れる可能性があります。
 > ご理解とご支援に感謝します。
 
+> [!TIP]
+> **Building in Public**
+>
+> メンテナーが Jobdori を使い、oh-my-opencode をリアルタイムで開発・メンテナンスしています。Jobdori は OpenClaw をベースに大幅カスタマイズされた AI アシスタントです。
+> すべての機能開発、修正、Issue トリアージを Discord でライブでご覧いただけます。
+>
+> [![Building in Public](./.github/assets/building-in-public.png)](https://discord.gg/PUwSMR9XNk)
+>
+> [**→ #building-in-public で確認する**](https://discord.gg/PUwSMR9XNk)
+
+
 > [!NOTE]
 >
 > [![Sisyphus Labs - Sisyphus is the agent that codes like your team.](./.github/assets/sisyphuslabs.png?v=2)](https://sisyphuslabs.ai)

README.ko.md

@@ -4,6 +4,17 @@
 > 핵심 메인테이너 Q가 부상을 입어, 이번 주에는 이슈/PR 응답 및 릴리스가 지연될 수 있습니다.
 > 양해와 응원에 감사드립니다.
 
+> [!TIP]
+> **Building in Public**
+>
+> 메인테이너가 Jobdori를 통해 oh-my-opencode를 실시간으로 개발하고 있습니다. Jobdori는 OpenClaw를 기반으로 대폭 커스터마이징된 AI 어시스턴트입니다.
+> 모든 기능 개발, 버그 수정, 이슈 트리아지를 Discord에서 실시간으로 확인하세요.
+>
+> [![Building in Public](./.github/assets/building-in-public.png)](https://discord.gg/PUwSMR9XNk)
+>
+> [**→ #building-in-public에서 확인하기**](https://discord.gg/PUwSMR9XNk)
+
+
 > [!TIP]
 > 저희와 함께 하세요!
 >

README.md

@@ -1,8 +1,12 @@
-> [!WARNING]
+> [!TIP]
-> **TEMP NOTICE (This Week): Reduced Maintainer Availability**
+> **Building in Public**
 >
-> Core maintainer Q got injured, so issue/PR responses and releases may be delayed this week.
+> The maintainer builds and maintains oh-my-opencode in real-time with Jobdori, an AI assistant built on a heavily customized fork of OpenClaw.
-> Thank you for your patience and support.
+> Every feature, every fix, every issue triage — live in our Discord.
+>
+> [![Building in Public](./.github/assets/building-in-public.png)](https://discord.gg/PUwSMR9XNk)
+>
+> [**→ Watch it happen in #building-in-public**](https://discord.gg/PUwSMR9XNk)
 
 > [!NOTE]
 >
@@ -37,7 +41,7 @@
 <div align="center">
 
 [![GitHub Release](https://img.shields.io/github/v/release/code-yeongyu/oh-my-openagent?color=369eff&labelColor=black&logo=github&style=flat-square)](https://github.com/code-yeongyu/oh-my-openagent/releases)
-[![npm downloads](https://img.shields.io/npm/dt/oh-my-opencode?color=ff6b35&labelColor=black&style=flat-square)](https://www.npmjs.com/package/oh-my-opencode)
+[![npm downloads](https://img.shields.io/endpoint?url=https%3A%2F%2Fohmyopenagent.com%2Fapi%2Fnpm-downloads&style=flat-square)](https://www.npmjs.com/package/oh-my-opencode)
 [![GitHub Contributors](https://img.shields.io/github/contributors/code-yeongyu/oh-my-openagent?color=c4f042&labelColor=black&style=flat-square)](https://github.com/code-yeongyu/oh-my-openagent/graphs/contributors)
 [![GitHub Forks](https://img.shields.io/github/forks/code-yeongyu/oh-my-openagent?color=8ae8ff&labelColor=black&style=flat-square)](https://github.com/code-yeongyu/oh-my-openagent/network/members)
 [![GitHub Stars](https://img.shields.io/github/stars/code-yeongyu/oh-my-openagent?color=ffcb47&labelColor=black&style=flat-square)](https://github.com/code-yeongyu/oh-my-openagent/stargazers)
@@ -181,7 +185,7 @@ When Sisyphus delegates to a subagent, it doesn't pick a model. It picks a **cat
 | `quick`              | Single-file changes, typos         |
 | `ultrabrain`         | Hard logic, architecture decisions |
 
-Agent says what kind of work. Harness picks the right model. You touch nothing.
+Agent says what kind of work. Harness picks the right model. `ultrabrain` now routes to GPT-5.4 xhigh by default. You touch nothing.
 
 ### Claude Code Compatibility
 

README.ru.md

@@ -4,6 +4,17 @@
 > Ключевой мейнтейнер Q получил травму, поэтому на этой неделе ответы по issue/PR и релизы могут задерживаться.
 > Спасибо за терпение и поддержку.
 
+> [!TIP]
+> **Building in Public**
+>
+> Мейнтейнер разрабатывает и поддерживает oh-my-opencode в режиме реального времени с помощью Jobdori — ИИ-ассистента на базе глубоко кастомизированной версии OpenClaw.
+> Каждая фича, каждый фикс, каждый триаж issue — в прямом эфире в нашем Discord.
+>
+> [![Building in Public](./.github/assets/building-in-public.png)](https://discord.gg/PUwSMR9XNk)
+>
+> [**→ Смотрите в #building-in-public**](https://discord.gg/PUwSMR9XNk)
+
+
 > [!NOTE]
 >
 > [![Sisyphus Labs - Sisyphus is the agent that codes like your team.](./.github/assets/sisyphuslabs.png?v=2)](https://sisyphuslabs.ai)

README.zh-cn.md

@@ -4,6 +4,17 @@
 > 核心维护者 Q 因受伤，本周 issue/PR 回复和发布可能会延迟。
 > 感谢你的耐心与支持。
 
+> [!TIP]
+> **Building in Public**
+>
+> 维护者正在使用 Jobdori 实时开发和维护 oh-my-opencode。Jobdori 是基于 OpenClaw 深度定制的 AI 助手。
+> 每个功能开发、每次修复、每次 Issue 分类，都在 Discord 上实时进行。
+>
+> [![Building in Public](./.github/assets/building-in-public.png)](https://discord.gg/PUwSMR9XNk)
+>
+> [**→ 在 #building-in-public 频道中查看**](https://discord.gg/PUwSMR9XNk)
+
+
 > [!NOTE]
 >
 > [![Sisyphus Labs - Sisyphus is the agent that codes like your team.](./.github/assets/sisyphuslabs.png?v=2)](https://sisyphuslabs.ai)

benchmarks/bun.lock

@@ -1,62 +0,0 @@
-{
-  "lockfileVersion": 1,
-  "configVersion": 1,
-  "workspaces": {
-    "": {
-      "name": "hashline-edit-benchmark",
-      "dependencies": {
-        "@ai-sdk/openai": "^1.3.0",
-        "@friendliai/ai-provider": "^1.0.9",
-        "ai": "^6.0.94",
-        "zod": "^4.1.0",
-      },
-    },
-  },
-  "packages": {
-    "@ai-sdk/gateway": ["@ai-sdk/gateway@3.0.55", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.15", "@vercel/oidc": "3.1.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-7xMeTJnCjwRwXKVCiv4Ly4qzWvDuW3+W1WIV0X1EFu6W83d4mEhV9bFArto10MeTw40ewuDjrbrZd21mXKohkw=="],
-
-    "@ai-sdk/openai": ["@ai-sdk/openai@1.3.24", "", { "dependencies": { "@ai-sdk/provider": "1.1.3", "@ai-sdk/provider-utils": "2.2.8" }, "peerDependencies": { "zod": "^3.0.0" } }, "sha512-GYXnGJTHRTZc4gJMSmFRgEQudjqd4PUN0ZjQhPwOAYH1yOAvQoG/Ikqs+HyISRbLPCrhbZnPKCNHuRU4OfpW0Q=="],
-
-    "@ai-sdk/openai-compatible": ["@ai-sdk/openai-compatible@2.0.30", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.15" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-iTjumHf1/u4NhjXYFn/aONM2GId3/o7J1Lp5ql8FCbgIMyRwrmanR5xy1S3aaVkfTscuDvLTzWiy1mAbGzK3nQ=="],
-
-    "@ai-sdk/provider": ["@ai-sdk/provider@1.1.3", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-qZMxYJ0qqX/RfnuIaab+zp8UAeJn/ygXXAffR5I4N0n1IrvA6qBsjc8hXLmBiMV2zoXlifkacF7sEFnYnjBcqg=="],
-
-    "@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@2.2.8", "", { "dependencies": { "@ai-sdk/provider": "1.1.3", "nanoid": "^3.3.8", "secure-json-parse": "^2.7.0" }, "peerDependencies": { "zod": "^3.23.8" } }, "sha512-fqhG+4sCVv8x7nFzYnFo19ryhAa3w096Kmc3hWxMQfW/TubPOmt3A6tYZhl4mUfQWWQMsuSkLrtjlWuXBVSGQA=="],
-
-    "@friendliai/ai-provider": ["@friendliai/ai-provider@1.1.4", "", { "dependencies": { "@ai-sdk/openai-compatible": "2.0.30", "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.15" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.12" } }, "sha512-9TU4B1QFqPhbkONjI5afCF7Ox4jOqtGg1xw8mA9QHZdtlEbZxU+mBNvMPlI5pU5kPoN6s7wkXmFmxpID+own1A=="],
-
-    "@opentelemetry/api": ["@opentelemetry/api@1.9.0", "", {}, "sha512-3giAOQvZiH5F9bMlMiv8+GSPMeqg0dbaeo58/0SlA9sxSqZhnUtxzX9/2FzyhS9sWQf5S0GJE0AKBrFqjpeYcg=="],
-
-    "@standard-schema/spec": ["@standard-schema/spec@1.1.0", "", {}, "sha512-l2aFy5jALhniG5HgqrD6jXLi/rUWrKvqN/qJx6yoJsgKhblVd+iqqU4RCXavm/jPityDo5TCvKMnpjKnOriy0w=="],
-
-    "@vercel/oidc": ["@vercel/oidc@3.1.0", "", {}, "sha512-Fw28YZpRnA3cAHHDlkt7xQHiJ0fcL+NRcIqsocZQUSmbzeIKRpwttJjik5ZGanXP+vlA4SbTg+AbA3bP363l+w=="],
-
-    "ai": ["ai@6.0.101", "", { "dependencies": { "@ai-sdk/gateway": "3.0.55", "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.15", "@opentelemetry/api": "1.9.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-Ur/NgbgOp1rdhyDiKDk6EOpSgd1g5ADlbcD1cjQJtQsnmhEngz3Rf8nK5JetDh0vnbLy2aEBpaQeL+zvLRWuaA=="],
-
-    "eventsource-parser": ["eventsource-parser@3.0.6", "", {}, "sha512-Vo1ab+QXPzZ4tCa8SwIHJFaSzy4R6SHf7BY79rFBDf0idraZWAkYrDjDj8uWaSm3S2TK+hJ7/t1CEmZ7jXw+pg=="],
-
-    "json-schema": ["json-schema@0.4.0", "", {}, "sha512-es94M3nTIfsEPisRafak+HDLfHXnKBhV3vU5eqPcS3flIWqcxJWgXHXiey3YrpaNsanY5ei1VoYEbOzijuq9BA=="],
-
-    "nanoid": ["nanoid@3.3.11", "", { "bin": { "nanoid": "bin/nanoid.cjs" } }, "sha512-N8SpfPUnUp1bK+PMYW8qSWdl9U+wwNWI4QKxOYDy9JAro3WMX7p2OeVRF9v+347pnakNevPmiHhNmZ2HbFA76w=="],
-
-    "secure-json-parse": ["secure-json-parse@2.7.0", "", {}, "sha512-6aU+Rwsezw7VR8/nyvKTx8QpWH9FrcYiXXlqC4z5d5XQBDRqtbfsRjnwGyqbi3gddNtWHuEk9OANUotL26qKUw=="],
-
-    "zod": ["zod@4.3.6", "", {}, "sha512-rftlrkhHZOcjDwkGlnUtZZkvaPHCsDATp4pGpuOOMDaTdDDXF91wuVDJoWoPsKX/3YPQ5fHuF3STjcYyKr+Qhg=="],
-
-    "@ai-sdk/gateway/@ai-sdk/provider": ["@ai-sdk/provider@3.0.8", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-oGMAgGoQdBXbZqNG0Ze56CHjDZ1IDYOwGYxYjO5KLSlz5HiNQ9udIXsPZ61VWaHGZ5XW/jyjmr6t2xz2jGVwbQ=="],
-
-    "@ai-sdk/gateway/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.15", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.6" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-8XiKWbemmCbvNN0CLR9u3PQiet4gtEVIrX4zzLxnCj06AwsEDJwJVBbKrEI4t6qE8XRSIvU2irka0dcpziKW6w=="],
-
-    "@ai-sdk/openai-compatible/@ai-sdk/provider": ["@ai-sdk/provider@3.0.8", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-oGMAgGoQdBXbZqNG0Ze56CHjDZ1IDYOwGYxYjO5KLSlz5HiNQ9udIXsPZ61VWaHGZ5XW/jyjmr6t2xz2jGVwbQ=="],
-
-    "@ai-sdk/openai-compatible/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.15", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.6" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-8XiKWbemmCbvNN0CLR9u3PQiet4gtEVIrX4zzLxnCj06AwsEDJwJVBbKrEI4t6qE8XRSIvU2irka0dcpziKW6w=="],
-
-    "@friendliai/ai-provider/@ai-sdk/provider": ["@ai-sdk/provider@3.0.8", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-oGMAgGoQdBXbZqNG0Ze56CHjDZ1IDYOwGYxYjO5KLSlz5HiNQ9udIXsPZ61VWaHGZ5XW/jyjmr6t2xz2jGVwbQ=="],
-
-    "@friendliai/ai-provider/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.15", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.6" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-8XiKWbemmCbvNN0CLR9u3PQiet4gtEVIrX4zzLxnCj06AwsEDJwJVBbKrEI4t6qE8XRSIvU2irka0dcpziKW6w=="],
-
-    "ai/@ai-sdk/provider": ["@ai-sdk/provider@3.0.8", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-oGMAgGoQdBXbZqNG0Ze56CHjDZ1IDYOwGYxYjO5KLSlz5HiNQ9udIXsPZ61VWaHGZ5XW/jyjmr6t2xz2jGVwbQ=="],
-
-    "ai/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.15", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.6" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-8XiKWbemmCbvNN0CLR9u3PQiet4gtEVIrX4zzLxnCj06AwsEDJwJVBbKrEI4t6qE8XRSIvU2irka0dcpziKW6w=="],
-  }
-}

benchmarks/package.json

@@ -1,19 +0,0 @@
-{
-  "name": "hashline-edit-benchmark",
-  "version": "0.1.0",
-  "private": true,
-  "type": "module",
-  "description": "Hashline edit tool benchmark using Vercel AI SDK with FriendliAI provider",
-  "scripts": {
-    "bench:basic": "bun run test-edit-ops.ts",
-    "bench:edge": "bun run test-edge-cases.ts",
-    "bench:multi": "bun run test-multi-model.ts",
-    "bench:all": "bun run bench:basic && bun run bench:edge"
-  },
-  "dependencies": {
-    "ai": "^6.0.94",
-    "@ai-sdk/openai": "^1.3.0",
-    "@friendliai/ai-provider": "^1.0.9",
-    "zod": "^4.1.0"
-  }
-}

bun.lock

@@ -5,13 +5,13 @@
     "": {
       "name": "oh-my-opencode",
       "dependencies": {
-        "@ast-grep/cli": "^0.40.0",
+        "@ast-grep/cli": "^0.41.1",
-        "@ast-grep/napi": "^0.40.0",
+        "@ast-grep/napi": "^0.41.1",
         "@clack/prompts": "^0.11.0",
         "@code-yeongyu/comment-checker": "^0.7.0",
         "@modelcontextprotocol/sdk": "^1.25.2",
-        "@opencode-ai/plugin": "^1.2.16",
+        "@opencode-ai/plugin": "^1.2.24",
-        "@opencode-ai/sdk": "^1.2.17",
+        "@opencode-ai/sdk": "^1.2.24",
         "commander": "^14.0.2",
         "detect-libc": "^2.0.0",
         "diff": "^8.0.3",
@@ -25,7 +25,7 @@
       "devDependencies": {
         "@types/js-yaml": "^4.0.9",
         "@types/picomatch": "^3.0.2",
-        "bun-types": "1.3.6",
+        "bun-types": "1.3.10",
         "typescript": "^5.7.3",
       },
       "optionalDependencies": {
@@ -49,44 +49,44 @@
     "@code-yeongyu/comment-checker",
   ],
   "overrides": {
-    "@opencode-ai/sdk": "^1.2.17",
+    "@opencode-ai/sdk": "^1.2.24",
   },
   "packages": {
-    "@ast-grep/cli": ["@ast-grep/cli@0.40.5", "", { "dependencies": { "detect-libc": "2.1.2" }, "optionalDependencies": { "@ast-grep/cli-darwin-arm64": "0.40.5", "@ast-grep/cli-darwin-x64": "0.40.5", "@ast-grep/cli-linux-arm64-gnu": "0.40.5", "@ast-grep/cli-linux-x64-gnu": "0.40.5", "@ast-grep/cli-win32-arm64-msvc": "0.40.5", "@ast-grep/cli-win32-ia32-msvc": "0.40.5", "@ast-grep/cli-win32-x64-msvc": "0.40.5" }, "bin": { "sg": "sg", "ast-grep": "ast-grep" } }, "sha512-yVXL7Gz0WIHerQLf+MVaVSkhIhidtWReG5akNVr/JS9OVCVkSdz7gWm7H8jVv2M9OO1tauuG76K3UaRGBPu5lQ=="],
+    "@ast-grep/cli": ["@ast-grep/cli@0.41.1", "", { "dependencies": { "detect-libc": "2.1.2" }, "optionalDependencies": { "@ast-grep/cli-darwin-arm64": "0.41.1", "@ast-grep/cli-darwin-x64": "0.41.1", "@ast-grep/cli-linux-arm64-gnu": "0.41.1", "@ast-grep/cli-linux-x64-gnu": "0.41.1", "@ast-grep/cli-win32-arm64-msvc": "0.41.1", "@ast-grep/cli-win32-ia32-msvc": "0.41.1", "@ast-grep/cli-win32-x64-msvc": "0.41.1" }, "bin": { "sg": "sg", "ast-grep": "ast-grep" } }, "sha512-6oSuzF1Ra0d9jdcmflRIR1DHcicI7TYVxaaV/hajV51J49r6C+1BA2H9G+e47lH4sDEXUS9KWLNGNvXa/Gqs5A=="],
 
-    "@ast-grep/cli-darwin-arm64": ["@ast-grep/cli-darwin-arm64@0.40.5", "", { "os": "darwin", "cpu": "arm64" }, "sha512-T9CzwJ1GqQhnANdsu6c7iT1akpvTVMK+AZrxnhIPv33Ze5hrXUUkqan+j4wUAukRJDqU7u94EhXLSLD+5tcJ8g=="],
+    "@ast-grep/cli-darwin-arm64": ["@ast-grep/cli-darwin-arm64@0.41.1", "", { "os": "darwin", "cpu": "arm64" }, "sha512-30lrXtyDB+16WS89Bk8sufA5TVUczyQye4PoIYLxZr+PRbPW7thpxHwBwGWL6QvPvUtlElrCe4seA1CEwFxeFA=="],
 
-    "@ast-grep/cli-darwin-x64": ["@ast-grep/cli-darwin-x64@0.40.5", "", { "os": "darwin", "cpu": "x64" }, "sha512-ez9b2zKvXU8f4ghhjlqYvbx6tWCKJTuVlNVqDDfjqwwhGeiTYfnzMlSVat4ElYRMd21gLtXZIMy055v2f21Ztg=="],
+    "@ast-grep/cli-darwin-x64": ["@ast-grep/cli-darwin-x64@0.41.1", "", { "os": "darwin", "cpu": "x64" }, "sha512-jRft57aWRgqYgLXooWxS9Nx5mb5JJ/KQIwEqacWkcmDZEdEui7oG50//6y4/vU5WRcS1n6oB2Vs7WBvTh3/Ypg=="],
 
-    "@ast-grep/cli-linux-arm64-gnu": ["@ast-grep/cli-linux-arm64-gnu@0.40.5", "", { "os": "linux", "cpu": "arm64" }, "sha512-VXa2L1IEYD66AMb0GuG7VlMMbPmEGoJUySWDcwSZo/D9neiry3MJ41LQR5oTG2HyhIPBsf9umrXnmuRq66BviA=="],
+    "@ast-grep/cli-linux-arm64-gnu": ["@ast-grep/cli-linux-arm64-gnu@0.41.1", "", { "os": "linux", "cpu": "arm64" }, "sha512-1XUL+8u+Xs1FoM2W6F4v8pRa2aQQcp5CZXBG8uy9n8FhwsQtrhBclJ2Vr9g/zzswHQT1293mnP5TOk1wlYZq6w=="],
 
-    "@ast-grep/cli-linux-x64-gnu": ["@ast-grep/cli-linux-x64-gnu@0.40.5", "", { "os": "linux", "cpu": "x64" }, "sha512-GQC5162eIOWXR2eQQ6Knzg7/8Trp5E1ODJkaErf0IubdQrZBGqj5AAcQPcWgPbbnmktjIp0H4NraPpOJ9eJ22A=="],
+    "@ast-grep/cli-linux-x64-gnu": ["@ast-grep/cli-linux-x64-gnu@0.41.1", "", { "os": "linux", "cpu": "x64" }, "sha512-oSsbXzbcl4hnRAw7b1bTFZapx9s+O8ToJJKI44oJAb7xKIG3Rubn2IMBOFvMvjjWEEax8PpS2IocgdB8nUAcbA=="],
 
-    "@ast-grep/cli-win32-arm64-msvc": ["@ast-grep/cli-win32-arm64-msvc@0.40.5", "", { "os": "win32", "cpu": "arm64" }, "sha512-YiZdnQZsSlXQTMsZJop/Ux9MmUGfuRvC2x/UbFgrt5OBSYxND+yoiMc0WcA3WG+wU+tt4ZkB5HUea3r/IkOLYA=="],
+    "@ast-grep/cli-win32-arm64-msvc": ["@ast-grep/cli-win32-arm64-msvc@0.41.1", "", { "os": "win32", "cpu": "arm64" }, "sha512-jTMNqjXnQUhInMB1X06sxWZJv/6pd4/iYSyk8RR5kdulnuNzoGEB9KYbm6ojxktPtMfZpb+7eShQLqqy/dG6Ag=="],
 
-    "@ast-grep/cli-win32-ia32-msvc": ["@ast-grep/cli-win32-ia32-msvc@0.40.5", "", { "os": "win32", "cpu": "ia32" }, "sha512-MHkCxCITVTr8sY9CcVqNKbfUzMa3Hc6IilGXad0Clnw2vNmPfWqSky+hU/UTerr5YHWwWfAVURH7ANZgirtx0Q=="],
+    "@ast-grep/cli-win32-ia32-msvc": ["@ast-grep/cli-win32-ia32-msvc@0.41.1", "", { "os": "win32", "cpu": "ia32" }, "sha512-mCTyr6/KQneKk0iYaWup4ywW5buNcFqL6TrJVfU0tkd38fu/RtJ5zywr978vVvFxsY+urRU0qkrmtQqXQNwDFA=="],
 
-    "@ast-grep/cli-win32-x64-msvc": ["@ast-grep/cli-win32-x64-msvc@0.40.5", "", { "os": "win32", "cpu": "x64" }, "sha512-/MJ5un7yxlClaaxou9eYl+Kr2xr/yTtYtTq5aLBWjPWA6dmmJ1nAJgx5zKHVuplFXFBrFDQk3paEgAETMTGcrA=="],
+    "@ast-grep/cli-win32-x64-msvc": ["@ast-grep/cli-win32-x64-msvc@0.41.1", "", { "os": "win32", "cpu": "x64" }, "sha512-AUbR67UKWsfgyy3SWQq258ZB0xSlaAe15Gl5hPu5tbUu4HTt6rKrUCTEEubYgbNdPPZWtxjobjFjMsDTWfnrug=="],
 
-    "@ast-grep/napi": ["@ast-grep/napi@0.40.5", "", { "optionalDependencies": { "@ast-grep/napi-darwin-arm64": "0.40.5", "@ast-grep/napi-darwin-x64": "0.40.5", "@ast-grep/napi-linux-arm64-gnu": "0.40.5", "@ast-grep/napi-linux-arm64-musl": "0.40.5", "@ast-grep/napi-linux-x64-gnu": "0.40.5", "@ast-grep/napi-linux-x64-musl": "0.40.5", "@ast-grep/napi-win32-arm64-msvc": "0.40.5", "@ast-grep/napi-win32-ia32-msvc": "0.40.5", "@ast-grep/napi-win32-x64-msvc": "0.40.5" } }, "sha512-hJA62OeBKUQT68DD2gDyhOqJxZxycqg8wLxbqjgqSzYttCMSDL9tiAQ9abgekBYNHudbJosm9sWOEbmCDfpX2A=="],
+    "@ast-grep/napi": ["@ast-grep/napi@0.41.1", "", { "optionalDependencies": { "@ast-grep/napi-darwin-arm64": "0.41.1", "@ast-grep/napi-darwin-x64": "0.41.1", "@ast-grep/napi-linux-arm64-gnu": "0.41.1", "@ast-grep/napi-linux-arm64-musl": "0.41.1", "@ast-grep/napi-linux-x64-gnu": "0.41.1", "@ast-grep/napi-linux-x64-musl": "0.41.1", "@ast-grep/napi-win32-arm64-msvc": "0.41.1", "@ast-grep/napi-win32-ia32-msvc": "0.41.1", "@ast-grep/napi-win32-x64-msvc": "0.41.1" } }, "sha512-OYQVWBbb43af2lTSCayMS7wsZ20nl+fw6LGVl/5zSuHTZRNfANknKLk3wMA4y7RIaAiIwrldAmI6GNZeIDRTkQ=="],
 
-    "@ast-grep/napi-darwin-arm64": ["@ast-grep/napi-darwin-arm64@0.40.5", "", { "os": "darwin", "cpu": "arm64" }, "sha512-2F072fGN0WTq7KI3okuEnkGJVEHLbi56Bw1H6NAMf7j2mJJeQWsRyGOMcyNnUXZDeNdvoMH0OB2a5wwUegY/nQ=="],
+    "@ast-grep/napi-darwin-arm64": ["@ast-grep/napi-darwin-arm64@0.41.1", "", { "os": "darwin", "cpu": "arm64" }, "sha512-sZHwg/oD6YB2y4VD8ZMeMHBq/ONil+mx+bB61YAiGQB+8UCMSFxJupvtNICB/BnIFqcPCVz/jCaSdbASLrbXQQ=="],
 
-    "@ast-grep/napi-darwin-x64": ["@ast-grep/napi-darwin-x64@0.40.5", "", { "os": "darwin", "cpu": "x64" }, "sha512-dJMidHZhhxuLBYNi6/FKI812jQ7wcFPSKkVPwviez2D+KvYagapUMAV/4dJ7FCORfguVk8Y0jpPAlYmWRT5nvA=="],
+    "@ast-grep/napi-darwin-x64": ["@ast-grep/napi-darwin-x64@0.41.1", "", { "os": "darwin", "cpu": "x64" }, "sha512-SL9hGB8sKvPnLUcigiDQrhohL7N4ujy1+t885kGcBkMXR73JT05OpPmvw0AWmg8l2iH1e5uNK/ZjnV/lSkynxQ=="],
 
-    "@ast-grep/napi-linux-arm64-gnu": ["@ast-grep/napi-linux-arm64-gnu@0.40.5", "", { "os": "linux", "cpu": "arm64" }, "sha512-nBRCbyoS87uqkaw4Oyfe5VO+SRm2B+0g0T8ME69Qry9ShMf41a2bTdpcQx9e8scZPogq+CTwDHo3THyBV71l9w=="],
+    "@ast-grep/napi-linux-arm64-gnu": ["@ast-grep/napi-linux-arm64-gnu@0.41.1", "", { "os": "linux", "cpu": "arm64" }, "sha512-mkNQpkm1jvnIdeRMnEWZ4Q0gNGApoNTMAoJRVmY11CkA4C/vIdNIjxj7UB61xV42Ng/A7Fw8mQUQuFos0lAKPQ=="],
 
-    "@ast-grep/napi-linux-arm64-musl": ["@ast-grep/napi-linux-arm64-musl@0.40.5", "", { "os": "linux", "cpu": "arm64" }, "sha512-/qKsmds5FMoaEj6FdNzepbmLMtlFuBLdrAn9GIWCqOIcVcYvM1Nka8+mncfeXB/MFZKOrzQsQdPTWqrrQzXLrA=="],
+    "@ast-grep/napi-linux-arm64-musl": ["@ast-grep/napi-linux-arm64-musl@0.41.1", "", { "os": "linux", "cpu": "arm64" }, "sha512-0G3cHyc+8A945aLie55bLZ+oaEBer0EFlyP/GlwRAx4nn5vGBct1hVTxSexWJ6AxnnRNPlN0mvswVwXiE7H7gA=="],
 
-    "@ast-grep/napi-linux-x64-gnu": ["@ast-grep/napi-linux-x64-gnu@0.40.5", "", { "os": "linux", "cpu": "x64" }, "sha512-DP4oDbq7f/1A2hRTFLhJfDFR6aI5mRWdEfKfHzRItmlKsR9WlcEl1qDJs/zX9R2EEtIDsSKRzuJNfJllY3/W8Q=="],
+    "@ast-grep/napi-linux-x64-gnu": ["@ast-grep/napi-linux-x64-gnu@0.41.1", "", { "os": "linux", "cpu": "x64" }, "sha512-+aNiCik3iTMtUrMp1k2yIMjby1U64EydTH1qotlx+fh8YvwrwwxZWct7NlurY3MILgT/WONSxhHKmL5NsbB4dw=="],
 
-    "@ast-grep/napi-linux-x64-musl": ["@ast-grep/napi-linux-x64-musl@0.40.5", "", { "os": "linux", "cpu": "x64" }, "sha512-BRZUvVBPUNpWPo6Ns8chXVzxHPY+k9gpsubGTHy92Q26ecZULd/dTkWWdnvfhRqttsSQ9Pe/XQdi5+hDQ6RYcg=="],
+    "@ast-grep/napi-linux-x64-musl": ["@ast-grep/napi-linux-x64-musl@0.41.1", "", { "os": "linux", "cpu": "x64" }, "sha512-rBrZSx5za3OliYcJcUrbLct+1+8oxh8ZEjYPiLCybe4FhspNKGM952g8a4sjgRuwbKS9BstYO9Fz+wthFnaFUQ=="],
 
-    "@ast-grep/napi-win32-arm64-msvc": ["@ast-grep/napi-win32-arm64-msvc@0.40.5", "", { "os": "win32", "cpu": "arm64" }, "sha512-y95zSEwc7vhxmcrcH0GnK4ZHEBQrmrszRBNQovzaciF9GUqEcCACNLoBesn4V47IaOp4fYgD2/EhGRTIBFb2Ug=="],
+    "@ast-grep/napi-win32-arm64-msvc": ["@ast-grep/napi-win32-arm64-msvc@0.41.1", "", { "os": "win32", "cpu": "arm64" }, "sha512-uNRHM3a1qFN0SECJDCEDVy1b0N75JNhJE2O/2BhDkDo0qM8kEewf9jRtG1fwpgZbMK2KoKvMHU/KQ73fWN44Zw=="],
 
-    "@ast-grep/napi-win32-ia32-msvc": ["@ast-grep/napi-win32-ia32-msvc@0.40.5", "", { "os": "win32", "cpu": "ia32" }, "sha512-K/u8De62iUnFCzVUs7FBdTZ2Jrgc5/DLHqjpup66KxZ7GIM9/HGME/O8aSoPkpcAeCD4TiTZ11C1i5p5H98hTg=="],
+    "@ast-grep/napi-win32-ia32-msvc": ["@ast-grep/napi-win32-ia32-msvc@0.41.1", "", { "os": "win32", "cpu": "ia32" }, "sha512-uNPQwGUBGIbCX+WhEIfYJf/VrS7o5+vJvT4MVEHI8aVJnpjcFsLrFI0hIv044OXxnleOo2HUvEmjOrub//at/Q=="],
 
-    "@ast-grep/napi-win32-x64-msvc": ["@ast-grep/napi-win32-x64-msvc@0.40.5", "", { "os": "win32", "cpu": "x64" }, "sha512-dqm5zg/o4Nh4VOQPEpMS23ot8HVd22gG0eg01t4CFcZeuzyuSgBlOL3N7xLbz3iH2sVkk7keuBwAzOIpTqziNQ=="],
+    "@ast-grep/napi-win32-x64-msvc": ["@ast-grep/napi-win32-x64-msvc@0.41.1", "", { "os": "win32", "cpu": "x64" }, "sha512-xFp68OCUEmWYcqoreZFaf2xwMhm/22Qf6bR2Qyn8WNVY9RF4m4+k5K+7Wn+n9xy0vHUPhtFd1So/SvuaqLHEoA=="],
 
     "@clack/core": ["@clack/core@0.5.0", "", { "dependencies": { "picocolors": "^1.0.0", "sisteransi": "^1.0.5" } }, "sha512-p3y0FIOwaYRUPRcMO7+dlmLh8PSRcrjuTndsiA0WAFbWES0mLZlrjVoBRZ9DzkPFJZG6KGkJmoEAY0ZcVWTkow=="],
 
@@ -98,9 +98,9 @@
 
     "@modelcontextprotocol/sdk": ["@modelcontextprotocol/sdk@1.27.1", "", { "dependencies": { "@hono/node-server": "^1.19.9", "ajv": "^8.17.1", "ajv-formats": "^3.0.1", "content-type": "^1.0.5", "cors": "^2.8.5", "cross-spawn": "^7.0.5", "eventsource": "^3.0.2", "eventsource-parser": "^3.0.0", "express": "^5.2.1", "express-rate-limit": "^8.2.1", "hono": "^4.11.4", "jose": "^6.1.3", "json-schema-typed": "^8.0.2", "pkce-challenge": "^5.0.0", "raw-body": "^3.0.0", "zod": "^3.25 || ^4.0", "zod-to-json-schema": "^3.25.1" }, "peerDependencies": { "@cfworker/json-schema": "^4.1.1" }, "optionalPeers": ["@cfworker/json-schema"] }, "sha512-sr6GbP+4edBwFndLbM60gf07z0FQ79gaExpnsjMGePXqFcSSb7t6iscpjk9DhFhwd+mTEQrzNafGP8/iGGFYaA=="],
 
-    "@opencode-ai/plugin": ["@opencode-ai/plugin@1.2.16", "", { "dependencies": { "@opencode-ai/sdk": "1.2.16", "zod": "4.1.8" } }, "sha512-9Kb7BQIC2P3oKCvI8K3thP5YP0vE7yLvcmBmgyACUIqc3e5UL6U+4umLpTvgQa2eQdjxtOXznuGTNwgcGMHUHg=="],
+    "@opencode-ai/plugin": ["@opencode-ai/plugin@1.2.24", "", { "dependencies": { "@opencode-ai/sdk": "1.2.24", "zod": "4.1.8" } }, "sha512-B3hw415D+2w6AtdRdvKWkuQVT0LXDWTdnAZhZC6gbd+UHh5O5DMmnZTe/YM8yK8ZZO9Dvo5rnV78TdDDYunJiw=="],
 
-    "@opencode-ai/sdk": ["@opencode-ai/sdk@1.2.17", "", {}, "sha512-HdeLeyJ2/Yl/NBHqw9pGFBnkIXuf0Id1kX1GMXDcnZwbJROUJ6TtrW/wLngTYW478E4CCm1jwknjxxmDuxzVMQ=="],
+    "@opencode-ai/sdk": ["@opencode-ai/sdk@1.2.24", "", {}, "sha512-MQamFkRl4B/3d6oIRLNpkYR2fcwet1V/ffKyOKJXWjtP/CT9PDJMtLpu6olVHjXKQi8zMNltwuMhv1QsNtRlZg=="],
 
     "@types/js-yaml": ["@types/js-yaml@4.0.9", "", {}, "sha512-k4MGaQl5TGo/iipqb2UDG2UwjXziSWkh0uysQelTlJpX1qGlpUZYm8PnO4DxG1qBomtJUdYJ6qR6xdIah10JLg=="],
 
@@ -118,7 +118,7 @@
 
     "body-parser": ["body-parser@2.2.2", "", { "dependencies": { "bytes": "^3.1.2", "content-type": "^1.0.5", "debug": "^4.4.3", "http-errors": "^2.0.0", "iconv-lite": "^0.7.0", "on-finished": "^2.4.1", "qs": "^6.14.1", "raw-body": "^3.0.1", "type-is": "^2.0.1" } }, "sha512-oP5VkATKlNwcgvxi0vM0p/D3n2C3EReYVX+DNYs5TjZFn/oQt2j+4sVJtSMr18pdRr8wjTcBl6LoV+FUwzPmNA=="],
 
-    "bun-types": ["bun-types@1.3.6", "", { "dependencies": { "@types/node": "*" } }, "sha512-OlFwHcnNV99r//9v5IIOgQ9Uk37gZqrNMCcqEaExdkVq3Avwqok1bJFmvGMCkCE0FqzdY8VMOZpfpR3lwI+CsQ=="],
+    "bun-types": ["bun-types@1.3.10", "", { "dependencies": { "@types/node": "*" } }, "sha512-tcpfCCl6XWo6nCVnpcVrxQ+9AYN1iqMIzgrSKYMB/fjLtV2eyAVEg7AxQJuCq/26R6HpKWykQXuSOq/21RYcbg=="],
 
     "bytes": ["bytes@3.1.2", "", {}, "sha512-/Nf7TyzTx6S3yRJObOAV7956r8cr2+Oj8AC5dt8wSP3BQAoeX58NoHyCU8P8zGkNXStjTSi6fzO6F0pBdcYbEg=="],
 

docs/guide/agent-model-matching.md

@@ -8,7 +8,7 @@ Think of AI models as developers on a team. Each has a different brain, differen
 
 This isn't a bug. It's the foundation of the entire system.
 
-Oh My OpenCode assigns each agent a model that matches its _working style_ — like building a team where each person is in the role that fits their personality.
+Oh My OpenAgent assigns each agent a model that matches its _working style_ — like building a team where each person is in the role that fits their personality.
 
 ### Sisyphus: The Sociable Lead
 
@@ -64,8 +64,8 @@ These agents have Claude-optimized prompts — long, detailed, mechanics-driven.
 
 | Agent        | Role              | Fallback Chain                         | Notes                                                                                             |
 | ------------ | ----------------- | -------------------------------------- | ------------------------------------------------------------------------------------------------- |
-| **Sisyphus** | Main orchestrator | Claude Opus → K2P5 → Kimi K2.5 → GPT-5.4 → GLM 5 → Big Pickle | Claude-family first. GPT-5.4 has dedicated prompt support. Kimi/GLM as intermediate fallbacks. |
+| **Sisyphus** | Main orchestrator | Claude Opus → opencode-go/kimi-k2.5 → K2P5 → Kimi K2.5 → GPT-5.4 → GLM-5 → Big Pickle | Claude-family first. GPT-5.4 has dedicated prompt support. Kimi available through multiple providers. |
-| **Metis**    | Plan gap analyzer | Claude Opus → GPT-5.4 → Gemini 3.1 Pro | Claude preferred, GPT acceptable fallback.                                                        |
+| **Metis**    | Plan gap analyzer | Claude Opus → GPT-5.4 → opencode-go/glm-5 → K2P5 | Claude preferred. GPT-5.4 as secondary before GLM-5 fallback.                                     |
 
 ### Dual-Prompt Agents → Claude preferred, GPT supported
 
@@ -73,8 +73,8 @@ These agents ship separate prompts for Claude and GPT families. They auto-detect
 
 | Agent          | Role              | Fallback Chain                         | Notes                                                                |
 | -------------- | ----------------- | -------------------------------------- | -------------------------------------------------------------------- |
-| **Prometheus** | Strategic planner | Claude Opus → GPT-5.4 → Gemini 3.1 Pro | Interview-mode planning. GPT prompt is compact and principle-driven. |
+| **Prometheus** | Strategic planner | Claude Opus → GPT-5.4 → opencode-go/glm-5 → Gemini 3.1 Pro | Interview-mode planning. GPT prompt is compact and principle-driven. |
-| **Atlas**      | Todo orchestrator | Claude Sonnet 4.6 → GPT-5.4            | Claude first, GPT-5.4 as the current fallback path.                  |
+| **Atlas**      | Todo orchestrator | Claude Sonnet → opencode-go/kimi-k2.5 → GPT-5.4 | Claude first, opencode-go as intermediate, GPT-5.4 as last resort.   |
 
 ### Deep Specialists → GPT
 
@@ -82,9 +82,9 @@ These agents are built for GPT's principle-driven style. Their prompts assume au
 
 | Agent          | Role                    | Fallback Chain                         | Notes                                            |
 | -------------- | ----------------------- | -------------------------------------- | ------------------------------------------------ |
-| **Hephaestus** | Autonomous deep worker  | GPT-5.3 Codex only                     | No fallback. Requires GPT access. The craftsman. |
+| **Hephaestus** | Autonomous deep worker  | GPT-5.3 Codex → GPT-5.4 (Copilot)     | Requires GPT access. GPT-5.4 via Copilot as fallback. The craftsman. |
-| **Oracle**     | Architecture consultant | GPT-5.4 → Gemini 3.1 Pro → Claude Opus | Read-only high-IQ consultation.                  |
+| **Oracle**     | Architecture consultant | GPT-5.4 → Gemini 3.1 Pro → Claude Opus → opencode-go/glm-5 | Read-only high-IQ consultation.                  |
-| **Momus**      | Ruthless reviewer       | GPT-5.4 → Claude Opus → Gemini 3.1 Pro | Verification and plan review.                    |
+| **Momus**      | Ruthless reviewer       | GPT-5.4 → Claude Opus → Gemini 3.1 Pro → opencode-go/glm-5 | Verification and plan review. GPT-5.4 uses xhigh variant. |
 
 ### Utility Runners → Speed over Intelligence
 
@@ -92,9 +92,10 @@ These agents do grep, search, and retrieval. They intentionally use the fastest,
 
 | Agent                 | Role               | Fallback Chain                                 | Notes                                                 |
 | --------------------- | ------------------ | ---------------------------------------------- | ----------------------------------------------------- |
-| **Explore**           | Fast codebase grep | Grok Code Fast → MiniMax → Haiku → GPT-5-Nano  | Speed is everything. Fire 10 in parallel.             |
+| **Explore**           | Fast codebase grep | Grok Code Fast → opencode-go/minimax-m2.7-highspeed → MiniMax M2.7 → Haiku → GPT-5-Nano | Speed is everything. Fire 10 in parallel.             |
-| **Librarian**         | Docs/code search   | Gemini Flash → MiniMax → Big Pickle            | Doc retrieval doesn't need deep reasoning.            |
+| **Librarian**         | Docs/code search   | opencode-go/minimax-m2.7 → MiniMax M2.7-highspeed → Haiku → GPT-5-Nano                  | Doc retrieval doesn't need deep reasoning.            |
-| **Multimodal Looker** | Vision/screenshots | GPT-5.3 Codex → K2P5 → Gemini Flash → GLM-4.6v | Uses the first available multimodal-capable fallback. |
+| **Multimodal Looker** | Vision/screenshots | GPT-5.4 → opencode-go/kimi-k2.5 → GLM-4.6v → GPT-5-Nano                                 | Uses the first available multimodal-capable fallback. |
+| **Sisyphus-Junior**   | Category executor  | Claude Sonnet → opencode-go/kimi-k2.5 → GPT-5.4 → MiniMax M2.7 → Big Pickle              | Handles delegated category tasks. Sonnet-tier default. |
 
 ---
 
@@ -119,8 +120,8 @@ Principle-driven, explicit reasoning, deep technical capability. Best for agents
 | Model             | Strengths                                                                                       |
 | ----------------- | ----------------------------------------------------------------------------------------------- |
 | **GPT-5.3 Codex** | Deep coding powerhouse. Autonomous exploration. Required for Hephaestus.                        |
-| **GPT-5.4**       | High intelligence, strategic reasoning. Default for Oracle.                                     |
+| **GPT-5.4**       | High intelligence, strategic reasoning. Default for Oracle, Momus, and a key fallback for Prometheus / Atlas. Uses xhigh variant for Momus. |
-| **GPT-5.4**       | Strong principle-driven reasoning. Default for Momus and a key fallback for Prometheus / Atlas. |
+| **GPT-5.4 Mini**  | Fast + strong reasoning. Good for lightweight autonomous tasks. Default for quick category. |
 | **GPT-5-Nano**    | Ultra-cheap, fast. Good for simple utility tasks.                                               |
 
 ### Other Models
@@ -130,11 +131,32 @@ Principle-driven, explicit reasoning, deep technical capability. Best for agents
 | **Gemini 3.1 Pro**   | Excels at visual/frontend tasks. Different reasoning style. Default for `visual-engineering` and `artistry`. |
 | **Gemini 3 Flash**   | Fast. Good for doc search and light tasks.                                                                   |
 | **Grok Code Fast 1** | Blazing fast code grep. Default for Explore agent.                                                           |
-| **MiniMax M2.5**     | Fast and smart. Good for utility tasks and search/retrieval.                                                 |
+| **MiniMax M2.7**     | Fast and smart. Good for utility tasks and search/retrieval. Upgraded from M2.5 with better reasoning.       |
+| **MiniMax M2.7 Highspeed** | Ultra-fast variant. Optimized for latency-sensitive tasks like codebase grep.                           |
+
+### OpenCode Go
+
+A premium subscription tier ($10/month) that provides reliable access to Chinese frontier models through OpenCode's infrastructure.
+
+**Available Models:**
+
+| Model                    | Use Case                                                              |
+| ------------------------ | --------------------------------------------------------------------- |
+| **opencode-go/kimi-k2.5** | Vision-capable, Claude-like reasoning. Used by Sisyphus, Atlas, Sisyphus-Junior, Multimodal Looker. |
+| **opencode-go/glm-5**     | Text-only orchestration model. Used by Oracle, Prometheus, Metis, Momus.                           |
+| **opencode-go/minimax-m2.7** | Ultra-cheap, fast responses. Used by Librarian, Explore, Atlas, Sisyphus-Junior for utility work.   |
+
+**When It Gets Used:**
+
+OpenCode Go models appear in fallback chains as intermediate options. They bridge the gap between premium Claude access and free-tier alternatives. The system tries OpenCode Go models before falling back to free tiers (MiniMax M2.7-highspeed, Big Pickle) or GPT alternatives.
+
+**Go-Only Scenarios:**
+
+Some model identifiers like `k2p5` (paid Kimi K2.5) and `glm-5` may only be available through OpenCode Go subscription in certain regions. When configured with these short identifiers, the system resolves them through the opencode-go provider first.
 
 ### About Free-Tier Fallbacks
 
-You may see model names like `kimi-k2.5-free`, `minimax-m2.5-free`, or `big-pickle` (GLM 4.6) in the source code or logs. These are free-tier versions of the same model families, served through the OpenCode Zen provider. They exist as lower-priority entries in fallback chains.
+You may see model names like `kimi-k2.5-free`, `minimax-m2.7-highspeed`, or `big-pickle` (GLM 4.6) in the source code or logs. These are free-tier or speed-optimized versions of the same model families. They exist as lower-priority entries in fallback chains.
 
 You don't need to configure them. The system includes them so it degrades gracefully when you don't have every paid subscription. If you have the paid version, the paid version is always preferred.
 
@@ -146,14 +168,14 @@ When agents delegate work, they don't pick a model name — they pick a **catego
 
 | Category             | When Used                  | Fallback Chain                               |
 | -------------------- | -------------------------- | -------------------------------------------- |
-| `visual-engineering` | Frontend, UI, CSS, design  | Gemini 3.1 Pro → GLM 5 → Claude Opus         |
+| `visual-engineering` | Frontend, UI, CSS, design  | Gemini 3.1 Pro → GLM 5 → Claude Opus → opencode-go/glm-5 → K2P5 |
-| `ultrabrain`         | Maximum reasoning needed   | GPT-5.3 Codex → Gemini 3.1 Pro → Claude Opus |
+| `ultrabrain`         | Maximum reasoning needed   | GPT-5.4 → Gemini 3.1 Pro → Claude Opus → opencode-go/glm-5 |
 | `deep`               | Deep coding, complex logic | GPT-5.3 Codex → Claude Opus → Gemini 3.1 Pro |
 | `artistry`           | Creative, novel approaches | Gemini 3.1 Pro → Claude Opus → GPT-5.4       |
-| `quick`              | Simple, fast tasks         | Claude Haiku → Gemini Flash → GPT-5-Nano     |
+| `quick`              | Simple, fast tasks         | GPT-5.4 Mini → Claude Haiku → Gemini Flash → opencode-go/minimax-m2.7 → GPT-5-Nano |
-| `unspecified-high`   | General complex work       | GPT-5.4 → Claude Opus → GLM 5 → K2P5         |
+| `unspecified-high`   | General complex work       | Claude Opus → GPT-5.4 → GLM 5 → K2P5 → opencode-go/glm-5 → Kimi K2.5 |
-| `unspecified-low`    | General standard work      | Claude Sonnet → GPT-5.3 Codex → Gemini Flash |
+| `unspecified-low`    | General standard work      | Claude Sonnet → GPT-5.3 Codex → opencode-go/kimi-k2.5 → Gemini Flash |
-| `writing`            | Text, docs, prose          | Gemini Flash → Claude Sonnet                 |
+| `writing`            | Text, docs, prose          | Gemini Flash → opencode-go/kimi-k2.5 → Claude Sonnet |
 
 See the [Orchestration System Guide](./orchestration.md) for how agents dispatch tasks to categories.
 
@@ -165,7 +187,7 @@ See the [Orchestration System Guide](./orchestration.md) for how agents dispatch
 
 ```jsonc
 {
-  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json",
+  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-openagent.schema.json",
 
   "agents": {
     // Main orchestrator: Claude Opus or Kimi K2.5 work best
@@ -190,7 +212,7 @@ See the [Orchestration System Guide](./orchestration.md) for how agents dispatch
   "categories": {
     "quick": { "model": "opencode/gpt-5-nano" },
     "unspecified-low": { "model": "anthropic/claude-sonnet-4-6" },
-    "unspecified-high": { "model": "openai/gpt-5.4-high" },
+    "unspecified-high": { "model": "anthropic/claude-opus-4-6", "variant": "max" },
     "visual-engineering": {
       "model": "google/gemini-3.1-pro",
       "variant": "high",

docs/guide/installation.md

@@ -5,7 +5,7 @@
 Paste this into your llm agent session:
 
 ```
-Install and configure oh-my-opencode by following the instructions here:
+Install and configure oh-my-openagent by following the instructions here:
 https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/refs/heads/dev/docs/guide/installation.md
 ```
 
@@ -14,8 +14,8 @@ https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/refs/heads/dev/do
 Run the interactive installer:
 
 ```bash
-bunx oh-my-opencode install # recommended
+bunx oh-my-openagent install # recommended
-npx oh-my-opencode install # alternative
+npx oh-my-openagent install # alternative
 ```
 
 > **Note**: The CLI ships with standalone binaries for all major platforms. No runtime (Bun/Node.js) is required for CLI execution after installation.
@@ -34,7 +34,7 @@ After you install it, you can read this [overview guide](./overview.md) to under
 > curl -fsSL https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/refs/heads/dev/docs/guide/installation.md
 > ```
 
-If you're an LLM Agent helping set up oh-my-opencode, Welcome! You'll now help your user to install the best agent harness, OhMyOpenCode. Tell user this with greetings, with test 'oMoMoMoMo...'
+If you're an LLM Agent helping set up oh-my-openagent, Welcome! You'll now help your user to install the best agent harness, OhMyOpenAgent. Tell user this with greetings, with test 'oMoMoMoMo...'
 
 follow these steps:
 
@@ -68,7 +68,12 @@ Ask the user these questions to determine CLI options:
    - If **yes** → `--zai-coding-plan=yes`
    - If **no** → `--zai-coding-plan=no` (default)
 
-**Provider Priority**: Native (anthropic/, openai/, google/) > GitHub Copilot > OpenCode Zen > Z.ai Coding Plan
+7. **Do you have an OpenCode Go subscription?**
+   - OpenCode Go is a $10/month subscription providing access to GLM-5, Kimi K2.5, and MiniMax M2.7 models
+   - If **yes** → `--opencode-go=yes`
+   - If **no** → `--opencode-go=no` (default)
+
+**Provider Priority**: Native (anthropic/, openai/, google/) > Kimi for Coding > GitHub Copilot > OpenCode Go > OpenCode Zen > Z.ai Coding Plan
 
 MUST STRONGLY WARNING, WHEN USER SAID THEY DON'T HAVE CLAUDE SUBSCRIPTION, SISYPHUS AGENT MIGHT NOT WORK IDEALLY.
 
@@ -91,18 +96,19 @@ Spawn a subagent to handle installation and report back - to save context.
 Based on user's answers, run the CLI installer with appropriate flags:
 
 ```bash
-bunx oh-my-opencode install --no-tui --claude=<yes|no|max20> --gemini=<yes|no> --copilot=<yes|no> [--openai=<yes|no>] [--opencode-zen=<yes|no>] [--zai-coding-plan=<yes|no>]
+bunx oh-my-openagent install --no-tui --claude=<yes|no|max20> --gemini=<yes|no> --copilot=<yes|no> [--openai=<yes|no>] [--opencode-go=<yes|no>] [--opencode-zen=<yes|no>] [--zai-coding-plan=<yes|no>]
 ```
 
 **Examples:**
 
-- User has all native subscriptions: `bunx oh-my-opencode install --no-tui --claude=max20 --openai=yes --gemini=yes --copilot=no`
+- User has all native subscriptions: `bunx oh-my-openagent install --no-tui --claude=max20 --openai=yes --gemini=yes --copilot=no`
-- User has only Claude: `bunx oh-my-opencode install --no-tui --claude=yes --gemini=no --copilot=no`
+- User has only Claude: `bunx oh-my-openagent install --no-tui --claude=yes --gemini=no --copilot=no`
-- User has Claude + OpenAI: `bunx oh-my-opencode install --no-tui --claude=yes --openai=yes --gemini=no --copilot=no`
+- User has Claude + OpenAI: `bunx oh-my-openagent install --no-tui --claude=yes --openai=yes --gemini=no --copilot=no`
-- User has only GitHub Copilot: `bunx oh-my-opencode install --no-tui --claude=no --gemini=no --copilot=yes`
+- User has only GitHub Copilot: `bunx oh-my-openagent install --no-tui --claude=no --gemini=no --copilot=yes`
-- User has Z.ai for Librarian: `bunx oh-my-opencode install --no-tui --claude=yes --gemini=no --copilot=no --zai-coding-plan=yes`
+- User has Z.ai for Librarian: `bunx oh-my-openagent install --no-tui --claude=yes --gemini=no --copilot=no --zai-coding-plan=yes`
-- User has only OpenCode Zen: `bunx oh-my-opencode install --no-tui --claude=no --gemini=no --copilot=no --opencode-zen=yes`
+- User has only OpenCode Zen: `bunx oh-my-openagent install --no-tui --claude=no --gemini=no --copilot=no --opencode-zen=yes`
-- User has no subscriptions: `bunx oh-my-opencode install --no-tui --claude=no --gemini=no --copilot=no`
+- User has OpenCode Go only: `bunx oh-my-openagent install --no-tui --claude=no --openai=no --gemini=no --copilot=no --opencode-go=yes`
+- User has no subscriptions: `bunx oh-my-openagent install --no-tui --claude=no --gemini=no --copilot=no`
 
 The CLI will:
 
@@ -114,7 +120,7 @@ The CLI will:
 
 ```bash
 opencode --version  # Should be 1.0.150 or higher
-cat ~/.config/opencode/opencode.json  # Should contain "oh-my-opencode" in plugin array
+cat ~/.config/opencode/opencode.json  # Should contain "oh-my-openagent" in plugin array
 ```
 
 ### Step 4: Configure Authentication
@@ -139,7 +145,7 @@ First, add the opencode-antigravity-auth plugin:
 
 ```json
 {
-  "plugin": ["oh-my-opencode", "opencode-antigravity-auth@latest"]
+  "plugin": ["oh-my-openagent", "opencode-antigravity-auth@latest"]
 }
 ```
 
@@ -148,9 +154,9 @@ First, add the opencode-antigravity-auth plugin:
 You'll also need full model settings in `opencode.json`.
 Read the [opencode-antigravity-auth documentation](https://github.com/NoeFabris/opencode-antigravity-auth), copy the full model configuration from the README, and merge carefully to avoid breaking the user's existing setup. The plugin now uses a **variant system** — models like `antigravity-gemini-3-pro` support `low`/`high` variants instead of separate `-low`/`-high` model entries.
 
-##### oh-my-opencode Agent Model Override
+##### oh-my-openagent Agent Model Override
 
-The `opencode-antigravity-auth` plugin uses different model names than the built-in Google auth. Override the agent models in `oh-my-opencode.json` (or `.opencode/oh-my-opencode.json`):
+The `opencode-antigravity-auth` plugin uses different model names than the built-in Google auth. Override the agent models in `oh-my-openagent.json` (or `.opencode/oh-my-openagent.json`):
 
 ```json
 {
@@ -170,7 +176,7 @@ The `opencode-antigravity-auth` plugin uses different model names than the built
 
 **Available models (Gemini CLI quota)**:
 
-- `google/gemini-2.5-flash`, `google/gemini-2.5-pro`, `google/gemini-3-flash-preview`, `google/gemini-3-pro-preview`
+- `google/gemini-2.5-flash`, `google/gemini-2.5-pro`, `google/gemini-3-flash-preview`, `google/gemini-3.1-pro-preview`
 
 > **Note**: Legacy tier-suffixed names like `google/antigravity-gemini-3-pro-high` still work but variants are recommended. Use `--variant=high` with the base model name instead.
 
@@ -195,11 +201,11 @@ GitHub Copilot is supported as a **fallback provider** when native providers are
 
 ##### Model Mappings
 
-When GitHub Copilot is the best available provider, oh-my-opencode uses these model assignments:
+When GitHub Copilot is the best available provider, oh-my-openagent uses these model assignments:
 
 | Agent         | Model                             |
 | ------------- | --------------------------------- |
-| **Sisyphus**  | `github-copilot/claude-opus-4-6`  |
+| **Sisyphus**  | `github-copilot/claude-opus-4.6`  |
 | **Oracle**    | `github-copilot/gpt-5.4`          |
 | **Explore**   | `github-copilot/grok-code-fast-1` |
 | **Librarian** | `github-copilot/gemini-3-flash`   |
@@ -221,7 +227,7 @@ If Z.ai is your main provider, the most important fallbacks are:
 
 #### OpenCode Zen
 
-OpenCode Zen provides access to `opencode/` prefixed models including `opencode/claude-opus-4-6`, `opencode/gpt-5.4`, `opencode/gpt-5.3-codex`, `opencode/gpt-5-nano`, `opencode/glm-5`, `opencode/big-pickle`, and `opencode/minimax-m2.5-free`.
+OpenCode Zen provides access to `opencode/` prefixed models including `opencode/claude-opus-4-6`, `opencode/gpt-5.4`, `opencode/gpt-5.3-codex`, `opencode/gpt-5-nano`, `opencode/glm-5`, `opencode/big-pickle`, and `opencode/minimax-m2.7-highspeed`.
 
 When OpenCode Zen is the best available provider (no native or Copilot), these models are used:
 
@@ -230,14 +236,14 @@ When OpenCode Zen is the best available provider (no native or Copilot), these m
 | **Sisyphus**  | `opencode/claude-opus-4-6`                           |
 | **Oracle**    | `opencode/gpt-5.4`                                   |
 | **Explore**   | `opencode/gpt-5-nano`                                |
-| **Librarian** | `opencode/minimax-m2.5-free` / `opencode/big-pickle` |
+| **Librarian** | `opencode/minimax-m2.7-highspeed` / `opencode/big-pickle` |
 
 ##### Setup
 
 Run the installer and select "Yes" for GitHub Copilot:
 
 ```bash
-bunx oh-my-opencode install
+bunx oh-my-openagent install
 # Select your subscriptions (Claude, ChatGPT, Gemini)
 # When prompted: "Do you have a GitHub Copilot subscription?" → Select "Yes"
 ```
@@ -245,7 +251,7 @@ bunx oh-my-opencode install
 Or use non-interactive mode:
 
 ```bash
-bunx oh-my-opencode install --no-tui --claude=no --openai=no --gemini=no --copilot=yes
+bunx oh-my-openagent install --no-tui --claude=no --openai=no --gemini=no --copilot=yes
 ```
 
 Then authenticate with GitHub:
@@ -257,7 +263,7 @@ opencode auth login
 
 ### Step 5: Understand Your Model Setup
 
-You've just configured oh-my-opencode. Here's what got set up and why.
+You've just configured oh-my-openagent. Here's what got set up and why.
 
 #### Model Families: What You're Working With
 
@@ -281,16 +287,17 @@ Not all models behave the same way. Understanding which models are "similar" hel
 | ----------------- | -------------------------------- | ------------------------------------------------- |
 | **GPT-5.3-codex** | openai, github-copilot, opencode | Deep coding powerhouse. Required for Hephaestus.  |
 | **GPT-5.4**       | openai, github-copilot, opencode | High intelligence. Default for Oracle.            |
+| **GPT-5.4 Mini**  | openai, github-copilot, opencode | Fast + strong reasoning. Default for quick category.     |
 | **GPT-5-Nano**    | opencode                         | Ultra-cheap, fast. Good for simple utility tasks. |
 
 **Different-Behavior Models**:
 
 | Model                 | Provider(s)                      | Notes                                                       |
 | --------------------- | -------------------------------- | ----------------------------------------------------------- |
-| **Gemini 3 Pro**      | google, github-copilot, opencode | Excels at visual/frontend tasks. Different reasoning style. |
+| **Gemini 3.1 Pro**    | google, github-copilot, opencode | Excels at visual/frontend tasks. Different reasoning style. |
 | **Gemini 3 Flash**    | google, github-copilot, opencode | Fast, good for doc search and light tasks.                  |
-| **MiniMax M2.5**      | venice                           | Fast and smart. Good for utility tasks.                     |
+| **MiniMax M2.7**      | venice, opencode-go              | Fast and smart. Good for utility tasks. Upgraded from M2.5. |
-| **MiniMax M2.5 Free** | opencode                         | Free-tier MiniMax. Fast for search/retrieval.               |
+| **MiniMax M2.7 Highspeed** | opencode                    | Ultra-fast MiniMax variant. Optimized for latency.          |
 
 **Speed-Focused Models**:
 
@@ -298,8 +305,8 @@ Not all models behave the same way. Understanding which models are "similar" hel
 | ----------------------- | ---------------------- | -------------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
 | **Grok Code Fast 1**    | github-copilot, venice | Very fast      | Optimized for code grep/search. Default for Explore.                                                                                          |
 | **Claude Haiku 4.5**    | anthropic, opencode    | Fast           | Good balance of speed and intelligence.                                                                                                       |
-| **MiniMax M2.5 (Free)** | opencode, venice       | Fast           | Smart for its speed class.                                                                                                                    |
+| **MiniMax M2.7 Highspeed** | opencode            | Very fast      | Ultra-fast MiniMax variant. Smart for its speed class.                                                                                        |
-| **GPT-5.3-codex-spark** | openai                 | Extremely fast | Blazing fast but compacts so aggressively that oh-my-opencode's context management doesn't work well with it. Not recommended for omo agents. |
+| **GPT-5.3-codex-spark** | openai                 | Extremely fast | Blazing fast but compacts so aggressively that oh-my-openagent's context management doesn't work well with it. Not recommended for omo agents. |
 
 #### What Each Agent Does and Which Model It Got
 
@@ -310,7 +317,7 @@ Based on your subscriptions, here's how the agents were configured:
 | Agent        | Role             | Default Chain                                   | What It Does                                                                             |
 | ------------ | ---------------- | ----------------------------------------------- | ---------------------------------------------------------------------------------------- |
 | **Sisyphus** | Main ultraworker | Opus (max) → Kimi K2.5 → GLM 5 → Big Pickle     | Primary coding agent. Orchestrates everything. **Never use GPT — no GPT prompt exists.** |
-| **Metis**    | Plan review      | Opus (max) → Kimi K2.5 → GPT-5.4 → Gemini 3 Pro | Reviews Prometheus plans for gaps.                                                       |
+| **Metis**    | Plan review      | Opus (max) → Kimi K2.5 → GPT-5.4 → Gemini 3.1 Pro | Reviews Prometheus plans for gaps.                                                       |
 
 **Dual-Prompt Agents** (auto-switch between Claude and GPT prompts):
 
@@ -320,7 +327,7 @@ Priority: **Claude > GPT > Claude-like models**
 
 | Agent          | Role              | Default Chain                                              | GPT Prompt?                                                      |
 | -------------- | ----------------- | ---------------------------------------------------------- | ---------------------------------------------------------------- |
-| **Prometheus** | Strategic planner | Opus (max) → **GPT-5.4 (high)** → Kimi K2.5 → Gemini 3 Pro | Yes — XML-tagged, principle-driven (~300 lines vs ~1,100 Claude) |
+| **Prometheus** | Strategic planner | Opus (max) → **GPT-5.4 (high)** → Kimi K2.5 → Gemini 3.1 Pro | Yes — XML-tagged, principle-driven (~300 lines vs ~1,100 Claude) |
 | **Atlas**      | Todo orchestrator | **Kimi K2.5** → Sonnet → GPT-5.4                           | Yes — GPT-optimized todo management                              |
 
 **GPT-Native Agents** (built for GPT, don't override to Claude):
@@ -328,8 +335,8 @@ Priority: **Claude > GPT > Claude-like models**
 | Agent          | Role                   | Default Chain                          | Notes                                                  |
 | -------------- | ---------------------- | -------------------------------------- | ------------------------------------------------------ |
 | **Hephaestus** | Deep autonomous worker | GPT-5.3-codex (medium) only            | "Codex on steroids." No fallback. Requires GPT access. |
-| **Oracle**     | Architecture/debugging | GPT-5.4 (high) → Gemini 3 Pro → Opus   | High-IQ strategic backup. GPT preferred.               |
+| **Oracle**     | Architecture/debugging | GPT-5.4 (high) → Gemini 3.1 Pro → Opus  | High-IQ strategic backup. GPT preferred.               |
-| **Momus**      | High-accuracy reviewer | GPT-5.4 (medium) → Opus → Gemini 3 Pro | Verification agent. GPT preferred.                     |
+| **Momus**      | High-accuracy reviewer | GPT-5.4 (medium) → Opus → Gemini 3.1 Pro | Verification agent. GPT preferred.                     |
 
 **Utility Agents** (speed over intelligence):
 
@@ -337,8 +344,8 @@ These agents do search, grep, and retrieval. They intentionally use fast, cheap
 
 | Agent                 | Role               | Default Chain                                                          | Design Rationale                                               |
 | --------------------- | ------------------ | ---------------------------------------------------------------------- | -------------------------------------------------------------- |
-| **Explore**           | Fast codebase grep | MiniMax M2.5 Free → Grok Code Fast → MiniMax M2.5 → Haiku → GPT-5-Nano | Speed is everything. Grok is blazing fast for grep.            |
+| **Explore**           | Fast codebase grep | Grok Code Fast → MiniMax M2.7-highspeed → MiniMax M2.7 → Haiku → GPT-5-Nano | Speed is everything. Grok is blazing fast for grep.            |
-| **Librarian**         | Docs/code search   | MiniMax M2.5 Free → Gemini Flash → Big Pickle                          | Entirely free-tier. Doc retrieval doesn't need deep reasoning. |
+| **Librarian**         | Docs/code search   | MiniMax M2.7 → MiniMax M2.7-highspeed → Haiku → GPT-5-Nano                   | Doc retrieval doesn't need deep reasoning. MiniMax is fast.    |
 | **Multimodal Looker** | Vision/screenshots | Kimi K2.5 → Kimi Free → Gemini Flash → GPT-5.4 → GLM-4.6v              | Kimi excels at multimodal understanding.                       |
 
 #### Why Different Models Need Different Prompts
@@ -358,7 +365,7 @@ This is why Prometheus and Atlas ship separate prompts per model family — they
 
 #### Custom Model Configuration
 
-If the user wants to override which model an agent uses, you can customize in `oh-my-opencode.json`:
+If the user wants to override which model an agent uses, you can customize in `oh-my-openagent.json`:
 
 ```jsonc
 {
@@ -400,7 +407,7 @@ GPT (5.3-codex, 5.2) > Claude Opus (decent fallback) > Gemini (acceptable)
 
 #### Provider Priority Chain
 
-When multiple providers are available, oh-my-opencode uses this priority:
+When multiple providers are available, oh-my-openagent uses this priority:
 
 ```
 Native (anthropic/, openai/, google/) > Kimi for Coding > GitHub Copilot > Venice > OpenCode Zen > Z.ai Coding Plan
@@ -418,7 +425,7 @@ Read this document again, think about you have done everything correctly.
 
 #### Say 'Congratulations! 🎉' to the user
 
-Say to user: Congratulations! 🎉 You have successfully set up OhMyOpenCode! Type `opencode` in your terminal to start using it.
+Say to user: Congratulations! 🎉 You have successfully set up OhMyOpenAgent! Type `opencode` in your terminal to start using it.
 
 #### Free advertising
 

docs/guide/orchestration.md

@@ -1,6 +1,6 @@
 # Orchestration System Guide
 
-Oh My OpenCode's orchestration system transforms a simple AI agent into a coordinated development team through **separation of planning and execution**.
+Oh My OpenAgent's orchestration system transforms a simple AI agent into a coordinated development team through **separation of planning and execution**.
 
 ---
 
@@ -296,12 +296,12 @@ task({ category: "quick", prompt: "..." }); // "Just get it done fast"
 | Category             | Model                  | When to Use                                                 |
 | -------------------- | ---------------------- | ----------------------------------------------------------- |
 | `visual-engineering` | Gemini 3.1 Pro         | Frontend, UI/UX, design, styling, animation                 |
-| `ultrabrain`         | GPT-5.3 Codex (xhigh)  | Deep logical reasoning, complex architecture decisions      |
+| `ultrabrain`         | GPT-5.4 (xhigh)        | Deep logical reasoning, complex architecture decisions      |
 | `artistry`           | Gemini 3.1 Pro (high)  | Highly creative or artistic tasks, novel ideas              |
-| `quick`              | Claude Haiku 4.5       | Trivial tasks - single file changes, typo fixes             |
+| `quick`              | GPT-5.4 Mini           | Trivial tasks - single file changes, typo fixes             |
 | `deep`               | GPT-5.3 Codex (medium) | Goal-oriented autonomous problem-solving, thorough research |
 | `unspecified-low`    | Claude Sonnet 4.6      | Tasks that don't fit other categories, low effort           |
-| `unspecified-high`   | GPT-5.4 (high)         | Tasks that don't fit other categories, high effort          |
+| `unspecified-high`   | Claude Opus 4.6 (max)  | Tasks that don't fit other categories, high effort          |
 | `writing`            | Gemini 3 Flash         | Documentation, prose, technical writing                     |
 
 ### Skills: Domain-Specific Instructions
@@ -475,7 +475,7 @@ Use the `ulw` keyword in Sisyphus when:
 
 ## Configuration
 
-You can control related features in `oh-my-opencode.json`:
+You can control related features in `oh-my-openagent.json`:
 
 ```jsonc
 {

docs/guide/overview.md

@@ -1,6 +1,6 @@
-# What Is Oh My OpenCode?
+# What Is Oh My OpenAgent?
 
-Oh My OpenCode is a multi-model agent orchestration harness for OpenCode. It transforms a single AI agent into a coordinated development team that actually ships code.
+Oh My OpenAgent is a multi-model agent orchestration harness for OpenCode. It transforms a single AI agent into a coordinated development team that actually ships code.
 
 Not locked to Claude. Not locked to OpenAI. Not locked to anyone.
 
@@ -15,7 +15,7 @@ Just better results, cheaper models, real orchestration.
 Paste this into your LLM agent session:
 
 ```
-Install and configure oh-my-opencode by following the instructions here:
+Install and configure oh-my-openagent by following the instructions here:
 https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/refs/heads/dev/docs/guide/installation.md
 ```
 
@@ -41,13 +41,13 @@ We used to call this "Claude Code on steroids." That was wrong.
 
 This isn't about making Claude Code better. It's about breaking free from the idea that one model, one provider, one way of working is enough. Anthropic wants you locked in. OpenAI wants you locked in. Everyone wants you locked in.
 
-Oh My OpenCode doesn't play that game. It orchestrates across models, picking the right brain for the right job. Claude for orchestration. GPT for deep reasoning. Gemini for frontend. Haiku for quick tasks. All working together, automatically.
+Oh My OpenAgent doesn't play that game. It orchestrates across models, picking the right brain for the right job. Claude for orchestration. GPT for deep reasoning. Gemini for frontend. GPT-5.4 Mini for quick tasks. All working together, automatically.
 
 ---
 
 ## How It Works: Agent Orchestration
 
-Instead of one agent doing everything, Oh My OpenCode uses **specialized agents that delegate to each other** based on task type.
+Instead of one agent doing everything, Oh My OpenAgent uses **specialized agents that delegate to each other** based on task type.
 
 **The Architecture:**
 
@@ -99,9 +99,9 @@ Use Hephaestus when you need deep architectural reasoning, complex debugging acr
 
 **Why this beats vanilla Codex CLI:**
 
-- **Multi-model orchestration.** Pure Codex is single-model. OmO routes different tasks to different models automatically. GPT for deep reasoning. Gemini for frontend. Haiku for speed. The right brain for the right job.
+- **Multi-model orchestration.** Pure Codex is single-model. OmO routes different tasks to different models automatically. GPT for deep reasoning. Gemini for frontend. GPT-5.4 Mini for speed. The right brain for the right job.
 - **Background agents.** Fire 5+ agents in parallel. Something Codex simply cannot do. While one agent writes code, another researches patterns, another checks documentation. Like a real dev team.
-- **Category system.** Tasks are routed by intent, not model name. `visual-engineering` gets Gemini. `ultrabrain` gets GPT-5.3 Codex. `quick` gets Haiku. No manual juggling.
+- **Category system.** Tasks are routed by intent, not model name. `visual-engineering` gets Gemini. `ultrabrain` gets GPT-5.4. `quick` gets GPT-5.4 Mini. No manual juggling.
 - **Accumulated wisdom.** Subagents learn from previous results. Conventions discovered in task 1 are passed to task 5. Mistakes made early aren't repeated. The system gets smarter as it works.
 
 ### Prometheus: The Strategic Planner
@@ -154,7 +154,7 @@ Use Prometheus for multi-day projects, critical production changes, complex refa
 
 ## Agent Model Matching
 
-Different agents work best with different models. Oh My OpenCode automatically assigns optimal models, but you can customize everything.
+Different agents work best with different models. Oh My OpenAgent automatically assigns optimal models, but you can customize everything.
 
 ### Default Configuration
 
@@ -168,7 +168,7 @@ You can override specific agents or categories in your config:
 
 ```jsonc
 {
-  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json",
+  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-openagent.schema.json",
 
   "agents": {
     // Main orchestrator: Claude Opus or Kimi K2.5 work best
@@ -193,13 +193,13 @@ You can override specific agents or categories in your config:
     },
 
     // General high-effort work
-    "unspecified-high": { "model": "openai/gpt-5.4", "variant": "high" },
+    "unspecified-high": { "model": "anthropic/claude-opus-4-6", "variant": "max" },
 
-    // Quick tasks: use the cheapest models
+    // Quick tasks: use GPT-5.4-mini (fast and cheap)
-    "quick": { "model": "anthropic/claude-haiku-4-5" },
+    "quick": { "model": "openai/gpt-5.4-mini" },
 
-    // Deep reasoning: GPT-5.3-codex
+    // Deep reasoning: GPT-5.4
-    "ultrabrain": { "model": "openai/gpt-5.3-codex", "variant": "xhigh" },
+    "ultrabrain": { "model": "openai/gpt-5.4", "variant": "xhigh" },
   },
 }
 ```
@@ -220,8 +220,8 @@ You can override specific agents or categories in your config:
 
 **Different-behavior models**:
 
-- Gemini 3 Pro — excels at visual/frontend tasks
+- Gemini 3.1 Pro — excels at visual/frontend tasks
-- MiniMax M2.5 — fast and smart for utility tasks
+- MiniMax M2.7 / M2.7-highspeed — fast and smart for utility tasks
 - Grok Code Fast 1 — optimized for code grep/search
 
 See the [Agent-Model Matching Guide](./agent-model-matching.md) for complete details on which models work best for each agent, safe vs dangerous overrides, and provider priority chains.
@@ -232,7 +232,7 @@ See the [Agent-Model Matching Guide](./agent-model-matching.md) for complete det
 
 Claude Code is good. But it's a single agent running a single model doing everything alone.
 
-Oh My OpenCode turns that into a coordinated team:
+Oh My OpenAgent turns that into a coordinated team:
 
 **Parallel execution.** Claude Code processes one thing at a time. OmO fires background agents in parallel — research, implementation, and verification happening simultaneously. Like having 5 engineers instead of 1.
 
@@ -246,7 +246,7 @@ Oh My OpenCode turns that into a coordinated team:
 
 **Discipline enforcement.** Todo enforcer yanks idle agents back to work. Comment checker strips AI slop. Ralph Loop keeps going until 100% done. The system doesn't let the agent slack off.
 
-**The fundamental advantage.** Models have different temperaments. Claude thinks deeply. GPT reasons architecturally. Gemini visualizes. Haiku moves fast. Single-model tools force you to pick one personality for all tasks. Oh My OpenCode leverages them all, routing by task type. This isn't a temporary hack — it's the only architecture that makes sense as models specialize further. The gap between multi-model orchestration and single-model limitation widens every month. We're betting on that future.
+**The fundamental advantage.** Models have different temperaments. Claude thinks deeply. GPT reasons architecturally. Gemini visualizes. Haiku moves fast. Single-model tools force you to pick one personality for all tasks. Oh My OpenAgent leverages them all, routing by task type. This isn't a temporary hack — it's the only architecture that makes sense as models specialize further. The gap between multi-model orchestration and single-model limitation widens every month. We're betting on that future.
 
 ---
 
@@ -256,7 +256,7 @@ Before acting on any request, Sisyphus classifies your true intent.
 
 Are you asking for research? Implementation? Investigation? A fix? The Intent Gate figures out what you actually want, not just the literal words you typed. This means the agent understands context, nuance, and the real goal behind your request.
 
-Claude Code doesn't have this. It takes your prompt and runs. Oh My OpenCode thinks first, then acts.
+Claude Code doesn't have this. It takes your prompt and runs. Oh My OpenAgent thinks first, then acts.
 
 ---
 

docs/manifesto.md

@@ -1,6 +1,6 @@
 # Manifesto
 
-The principles and philosophy behind Oh My OpenCode.
+The principles and philosophy behind Oh My OpenAgent.
 
 ---
 
@@ -20,7 +20,7 @@ When you find yourself:
 
 That's not "human-AI collaboration." That's the AI failing to do its job.
 
-**Oh My OpenCode is built on this premise**: Human intervention during agentic work is fundamentally a wrong signal. If the system is designed correctly, the agent should complete the work without requiring you to babysit it.
+**Oh My OpenAgent is built on this premise**: Human intervention during agentic work is fundamentally a wrong signal. If the system is designed correctly, the agent should complete the work without requiring you to babysit it.
 
 ---
 
@@ -144,7 +144,7 @@ Human Intent → Agent Execution → Verified Result
           (intervention only on true failure)
 ```
 
-Everything in Oh My OpenCode is designed to make this loop work:
+Everything in Oh My OpenAgent is designed to make this loop work:
 
 | Feature | Purpose |
 |---------|---------|

docs/model-capabilities-maintenance.md

@@ -0,0 +1,33 @@
+# Model Capabilities Maintenance
+
+This project treats model capability resolution as a layered system:
+
+1. runtime metadata from connected providers
+2. `models.dev` bundled/runtime snapshot data
+3. explicit compatibility aliases
+4. heuristic fallback as the last resort
+
+## Internal policy
+
+- Built-in OmO agent/category requirement models must use canonical model IDs.
+- Aliases exist only to preserve compatibility with historical OmO names or provider-specific decorations.
+- New decorated names like `-high`, `-low`, or `-thinking` should not be added to built-in requirements when a canonical model ID plus structured settings can express the same thing.
+- If a provider or config input still uses an alias, normalize it at the edge and continue internally with the canonical ID.
+
+## When adding an alias
+
+- Add the alias rule to `src/shared/model-capability-aliases.ts`.
+- Include a rationale for why the alias exists.
+- Add or update tests so the alias is covered explicitly.
+- Ensure the alias canonical target exists in the bundled `models.dev` snapshot.
+
+## Guardrails
+
+`bun run test:model-capabilities` enforces the following invariants:
+
+- exact alias targets must exist in the bundled snapshot
+- exact alias keys must not silently become canonical `models.dev` IDs
+- pattern aliases must not rewrite canonical snapshot IDs
+- built-in requirement models must stay canonical and snapshot-backed
+
+The scheduled `refresh-model-capabilities` workflow runs these guardrails before opening an automated snapshot refresh PR.

docs/reference/cli.md

@@ -1,15 +1,15 @@
 # CLI Reference
 
-Complete reference for the `oh-my-opencode` command-line interface.
+Complete reference for the `oh-my-openagent` command-line interface.
 
 ## Basic Usage
 
 ```bash
 # Display help
-bunx oh-my-opencode
+bunx oh-my-openagent
 
 # Or with npx
-npx oh-my-opencode
+npx oh-my-openagent
 ```
 
 ## Commands
@@ -27,20 +27,20 @@ npx oh-my-opencode
 
 ## install
 
-Interactive installation tool for initial Oh-My-OpenCode setup. Provides a TUI based on `@clack/prompts`.
+Interactive installation tool for initial Oh-My-OpenAgent setup. Provides a TUI based on `@clack/prompts`.
 
 ### Usage
 
 ```bash
-bunx oh-my-opencode install
+bunx oh-my-openagent install
 ```
 
 ### Installation Process
 
 1. **Provider Selection**: Choose your AI provider (Claude, ChatGPT, or Gemini)
 2. **API Key Input**: Enter the API key for your selected provider
-3. **Configuration File Creation**: Generates `opencode.json` or `oh-my-opencode.json` files
+3. **Configuration File Creation**: Generates `opencode.json` or `oh-my-openagent.json` files
-4. **Plugin Registration**: Automatically registers the oh-my-opencode plugin in OpenCode settings
+4. **Plugin Registration**: Automatically registers the oh-my-openagent plugin in OpenCode settings
 
 ### Options
 
@@ -53,12 +53,12 @@ bunx oh-my-opencode install
 
 ## doctor
 
-Diagnoses your environment to ensure Oh-My-OpenCode is functioning correctly. Performs 17+ health checks.
+Diagnoses your environment to ensure Oh-My-OpenAgent is functioning correctly. Performs 17+ health checks.
 
 ### Usage
 
 ```bash
-bunx oh-my-opencode doctor
+bunx oh-my-openagent doctor
 ```
 
 ### Diagnostic Categories
@@ -83,10 +83,10 @@ bunx oh-my-opencode doctor
 ### Example Output
 
 ```
-oh-my-opencode doctor
+oh-my-openagent doctor
 
 ┌──────────────────────────────────────────────────┐
-│  Oh-My-OpenCode Doctor                           │
+│  Oh-My-OpenAgent Doctor                           │
 └──────────────────────────────────────────────────┘
 
 Installation
@@ -94,7 +94,7 @@ Installation
   ✓ Plugin registered in opencode.json
 
 Configuration
-  ✓ oh-my-opencode.json is valid
+  ✓ oh-my-openagent.json is valid
   ⚠ categories.visual-engineering: using default model
 
 Authentication
@@ -119,7 +119,7 @@ Executes OpenCode sessions and monitors task completion.
 ### Usage
 
 ```bash
-bunx oh-my-opencode run [prompt]
+bunx oh-my-openagent run [prompt]
 ```
 
 ### Options
@@ -148,16 +148,16 @@ Manages OAuth 2.1 authentication for remote MCP servers.
 
 ```bash
 # Login to an OAuth-protected MCP server
-bunx oh-my-opencode mcp oauth login <server-name> --server-url https://api.example.com
+bunx oh-my-openagent mcp oauth login <server-name> --server-url https://api.example.com
 
 # Login with explicit client ID and scopes
-bunx oh-my-opencode mcp oauth login my-api --server-url https://api.example.com --client-id my-client --scopes "read,write"
+bunx oh-my-openagent mcp oauth login my-api --server-url https://api.example.com --client-id my-client --scopes "read,write"
 
 # Remove stored OAuth tokens
-bunx oh-my-opencode mcp oauth logout <server-name>
+bunx oh-my-openagent mcp oauth logout <server-name>
 
 # Check OAuth token status
-bunx oh-my-opencode mcp oauth status [server-name]
+bunx oh-my-openagent mcp oauth status [server-name]
 ```
 
 ### Options
@@ -178,8 +178,8 @@ Tokens are stored in `~/.config/opencode/mcp-oauth.json` with `0600` permissions
 
 The CLI searches for configuration files in the following locations (in priority order):
 
-1. **Project Level**: `.opencode/oh-my-opencode.json`
+1. **Project Level**: `.opencode/oh-my-openagent.json`
-2. **User Level**: `~/.config/opencode/oh-my-opencode.json`
+2. **User Level**: `~/.config/opencode/oh-my-openagent.json`
 
 ### JSONC Support
 
@@ -219,17 +219,17 @@ bun install -g opencode@latest
 
 ```bash
 # Reinstall plugin
-bunx oh-my-opencode install
+bunx oh-my-openagent install
 ```
 
 ### Doctor Check Failures
 
 ```bash
 # Diagnose with detailed information
-bunx oh-my-opencode doctor --verbose
+bunx oh-my-openagent doctor --verbose
 
 # Check specific category only
-bunx oh-my-opencode doctor --category authentication
+bunx oh-my-openagent doctor --category authentication
 ```
 
 ---
@@ -240,10 +240,10 @@ Use the `--no-tui` option for CI/CD environments.
 
 ```bash
 # Run doctor in CI environment
-bunx oh-my-opencode doctor --no-tui --json
+bunx oh-my-openagent doctor --no-tui --json
 
 # Save results to file
-bunx oh-my-opencode doctor --json > doctor-report.json
+bunx oh-my-openagent doctor --json > doctor-report.json
 ```
 
 ---

docs/reference/configuration.md

@@ -1,6 +1,6 @@
 # Configuration Reference
 
-Complete reference for `oh-my-opencode.jsonc` configuration. This document covers every available option with examples.
+Complete reference for `oh-my-openagent.jsonc` configuration. This document covers every available option with examples.
 
 ---
 
@@ -44,13 +44,13 @@ Complete reference for `oh-my-opencode.jsonc` configuration. This document cover
 
 Priority order (project overrides user):
 
-1. `.opencode/oh-my-opencode.jsonc` / `.opencode/oh-my-opencode.json`
+1. `.opencode/oh-my-openagent.jsonc` / `.opencode/oh-my-openagent.json`
 2. User config (`.jsonc` preferred over `.json`):
 
 | Platform    | Path                                      |
 | ----------- | ----------------------------------------- |
-| macOS/Linux | `~/.config/opencode/oh-my-opencode.jsonc` |
+| macOS/Linux | `~/.config/opencode/oh-my-openagent.jsonc` |
-| Windows     | `%APPDATA%\opencode\oh-my-opencode.jsonc` |
+| Windows     | `%APPDATA%\opencode\oh-my-openagent.jsonc` |
 
 JSONC supports `// line comments`, `/* block comments */`, and trailing commas.
 
@@ -58,11 +58,11 @@ Enable schema autocomplete:
 
 ```json
 {
-  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json"
+  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-openagent.schema.json"
 }
 ```
 
-Run `bunx oh-my-opencode install` for guided setup. Run `opencode models` to list available models.
+Run `bunx oh-my-openagent install` for guided setup. Run `opencode models` to list available models.
 
 ### Quick Start Example
 
@@ -70,7 +70,7 @@ Here's a practical starting configuration:
 
 ```jsonc
 {
-  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json",
+  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-openagent.schema.json",
 
   "agents": {
     // Main orchestrator: Claude Opus or Kimi K2.5 work best
@@ -100,7 +100,7 @@ Here's a practical starting configuration:
     "unspecified-low": { "model": "anthropic/claude-sonnet-4-6" },
 
     // unspecified-high — complex work
-    "unspecified-high": { "model": "openai/gpt-5.4-high" },
+    "unspecified-high": { "model": "anthropic/claude-opus-4-6", "variant": "max" },
 
     // writing — docs/prose
     "writing": { "model": "google/gemini-3-flash" },
@@ -225,12 +225,12 @@ Domain-specific model delegation used by the `task()` tool. When Sisyphus delega
 | Category             | Default Model                   | Description                                    |
 | -------------------- | ------------------------------- | ---------------------------------------------- |
 | `visual-engineering` | `google/gemini-3.1-pro` (high)  | Frontend, UI/UX, design, animation             |
-| `ultrabrain`         | `openai/gpt-5.3-codex` (xhigh)  | Deep logical reasoning, complex architecture   |
+| `ultrabrain`         | `openai/gpt-5.4` (xhigh)        | Deep logical reasoning, complex architecture   |
 | `deep`               | `openai/gpt-5.3-codex` (medium) | Autonomous problem-solving, thorough research  |
 | `artistry`           | `google/gemini-3.1-pro` (high)  | Creative/unconventional approaches             |
-| `quick`              | `anthropic/claude-haiku-4-5`    | Trivial tasks, typo fixes, single-file changes |
+| `quick`              | `openai/gpt-5.4-mini`           | Trivial tasks, typo fixes, single-file changes |
 | `unspecified-low`    | `anthropic/claude-sonnet-4-6`   | General tasks, low effort                      |
-| `unspecified-high`   | `openai/gpt-5.4` (high)         | General tasks, high effort                     |
+| `unspecified-high`   | `anthropic/claude-opus-4-6` (max) | General tasks, high effort                   |
 | `writing`            | `google/gemini-3-flash`         | Documentation, prose, technical writing        |
 
 > **Note**: Built-in defaults only apply if the category is present in your config. Otherwise the system default model is used.
@@ -270,8 +270,8 @@ Disable categories: `{ "disabled_categories": ["ultrabrain"] }`
 | **Sisyphus**          | `claude-opus-4-6`   | `claude-opus-4-6` → `glm-5` → `big-pickle`                                   |
 | **Hephaestus**        | `gpt-5.3-codex`     | `gpt-5.3-codex` → `gpt-5.4` (GitHub Copilot fallback)                        |
 | **oracle**            | `gpt-5.4`           | `gpt-5.4` → `gemini-3.1-pro` → `claude-opus-4-6`                             |
-| **librarian**         | `gemini-3-flash`    | `gemini-3-flash` → `minimax-m2.5-free` → `big-pickle`                        |
+| **librarian**         | `minimax-m2.7`      | `minimax-m2.7` → `minimax-m2.7-highspeed` → `claude-haiku-4-5` → `gpt-5-nano` |
-| **explore**           | `grok-code-fast-1`  | `grok-code-fast-1` → `minimax-m2.5-free` → `claude-haiku-4-5` → `gpt-5-nano` |
+| **explore**           | `grok-code-fast-1`  | `grok-code-fast-1` → `minimax-m2.7-highspeed` → `minimax-m2.7` → `claude-haiku-4-5` → `gpt-5-nano` |
 | **multimodal-looker** | `gpt-5.3-codex`     | `gpt-5.3-codex` → `k2p5` → `gemini-3-flash` → `glm-4.6v` → `gpt-5-nano`      |
 | **Prometheus**        | `claude-opus-4-6`   | `claude-opus-4-6` → `gpt-5.4` → `gemini-3.1-pro`                             |
 | **Metis**             | `claude-opus-4-6`   | `claude-opus-4-6` → `gpt-5.4` → `gemini-3.1-pro`                             |
@@ -283,15 +283,15 @@ Disable categories: `{ "disabled_categories": ["ultrabrain"] }`
 | Category               | Default Model       | Provider Priority                                              |
 | ---------------------- | ------------------- | -------------------------------------------------------------- |
 | **visual-engineering** | `gemini-3.1-pro`    | `gemini-3.1-pro` → `glm-5` → `claude-opus-4-6`                 |
-| **ultrabrain**         | `gpt-5.3-codex`     | `gpt-5.3-codex` → `gemini-3.1-pro` → `claude-opus-4-6`         |
+| **ultrabrain**         | `gpt-5.4`           | `gpt-5.4` → `gemini-3.1-pro` → `claude-opus-4-6`               |
 | **deep**               | `gpt-5.3-codex`     | `gpt-5.3-codex` → `claude-opus-4-6` → `gemini-3.1-pro`         |
 | **artistry**           | `gemini-3.1-pro`    | `gemini-3.1-pro` → `claude-opus-4-6` → `gpt-5.4`               |
-| **quick**              | `claude-haiku-4-5`  | `claude-haiku-4-5` → `gemini-3-flash` → `gpt-5-nano`           |
+| **quick**              | `gpt-5.4-mini`    | `gpt-5.4-mini` → `claude-haiku-4-5` → `gemini-3-flash` → `minimax-m2.7` → `gpt-5-nano` |
-| **unspecified-low**    | `claude-sonnet-4-6` | `claude-sonnet-4-6` → `gpt-5.3-codex` → `gemini-3-flash`       |
+| **unspecified-low**    | `claude-sonnet-4-6` | `claude-sonnet-4-6` → `gpt-5.3-codex` → `gemini-3-flash` → `minimax-m2.7` |
-| **unspecified-high**   | `gpt-5.4`           | `gpt-5.4` → `claude-opus-4-6` → `glm-5` → `k2p5` → `kimi-k2.5` |
+| **unspecified-high**   | `claude-opus-4-6`   | `claude-opus-4-6` → `gpt-5.4 (high)` → `glm-5` → `k2p5` → `kimi-k2.5` |
-| **writing**            | `gemini-3-flash`    | `gemini-3-flash` → `claude-sonnet-4-6`                         |
+| **writing**            | `gemini-3-flash`    | `gemini-3-flash` → `claude-sonnet-4-6` → `minimax-m2.7`        |
 
-Run `bunx oh-my-opencode doctor --verbose` to see effective model resolution for your config.
+Run `bunx oh-my-openagent doctor --verbose` to see effective model resolution for your config.
 
 ---
 
@@ -418,7 +418,7 @@ Disable built-in skills: `{ "disabled_skills": ["playwright"] }`
 Disable built-in hooks via `disabled_hooks`:
 
 ```json
-{ "disabled_hooks": ["comment-checker", "agent-usage-reminder"] }
+{ "disabled_hooks": ["comment-checker"] }
 ```
 
 Available hooks: `todo-continuation-enforcer`, `context-window-monitor`, `session-recovery`, `session-notification`, `comment-checker`, `grep-output-truncator`, `tool-output-truncator`, `directory-agents-injector`, `directory-readme-injector`, `empty-task-response-detector`, `think-mode`, `anthropic-context-window-limit-recovery`, `rules-injector`, `background-notification`, `auto-update-checker`, `startup-toast`, `keyword-detector`, `agent-usage-reminder`, `non-interactive-env`, `interactive-bash-session`, `compaction-context-injector`, `thinking-block-validator`, `claude-code-hooks`, `ralph-loop`, `preemptive-compaction`, `auto-slash-command`, `sisyphus-junior-notepad`, `no-sisyphus-gpt`, `start-work`, `runtime-fallback`

docs/reference/features.md

@@ -1,8 +1,8 @@
-# Oh-My-OpenCode Features Reference
+# Oh-My-OpenAgent Features Reference
 
 ## Agents
 
-Oh-My-OpenCode provides 11 specialized AI agents. Each has distinct expertise, optimized models, and tool permissions.
+Oh-My-OpenAgent provides 11 specialized AI agents. Each has distinct expertise, optimized models, and tool permissions.
 
 ### Core Agents
 
@@ -11,8 +11,8 @@ Oh-My-OpenCode provides 11 specialized AI agents. Each has distinct expertise, o
 | **Sisyphus**          | `claude-opus-4-6`  | The default orchestrator. Plans, delegates, and executes complex tasks using specialized subagents with aggressive parallel execution. Todo-driven workflow with extended thinking (32k budget). Fallback: `glm-5` → `big-pickle`.                                                                                                                               |
 | **Hephaestus**        | `gpt-5.3-codex`    | The Legitimate Craftsman. Autonomous deep worker inspired by AmpCode's deep mode. Goal-oriented execution with thorough research before action. Explores codebase patterns, completes tasks end-to-end without premature stopping. Named after the Greek god of forge and craftsmanship. Fallback: `gpt-5.4` on GitHub Copilot. Requires a GPT-capable provider. |
 | **Oracle**            | `gpt-5.4`          | Architecture decisions, code review, debugging. Read-only consultation with stellar logical reasoning and deep analysis. Inspired by AmpCode. Fallback: `gemini-3.1-pro` → `claude-opus-4-6`.                                                                                                                                                                    |
-| **Librarian**         | `gemini-3-flash`   | Multi-repo analysis, documentation lookup, OSS implementation examples. Deep codebase understanding with evidence-based answers. Fallback: `minimax-m2.5-free` → `big-pickle`.                                                                                                                                                                                   |
+| **Librarian**         | `minimax-m2.7`     | Multi-repo analysis, documentation lookup, OSS implementation examples. Deep codebase understanding with evidence-based answers. Fallback: `minimax-m2.7-highspeed` → `claude-haiku-4-5` → `gpt-5-nano`.                                                                                                                                                         |
-| **Explore**           | `grok-code-fast-1` | Fast codebase exploration and contextual grep. Fallback: `minimax-m2.5-free` → `claude-haiku-4-5` → `gpt-5-nano`.                                                                                                                                                                                                                                                |
+| **Explore**           | `grok-code-fast-1` | Fast codebase exploration and contextual grep. Fallback: `minimax-m2.7-highspeed` → `minimax-m2.7` → `claude-haiku-4-5` → `gpt-5-nano`.                                                                                                                                                                                                                          |
 | **Multimodal-Looker** | `gpt-5.3-codex`    | Visual content specialist. Analyzes PDFs, images, diagrams to extract information. Fallback: `k2p5` → `gemini-3-flash` → `glm-4.6v` → `gpt-5-nano`.                                                                                                                                                                                                              |
 
 ### Planning Agents
@@ -90,7 +90,7 @@ When running inside tmux:
 - Each pane shows agent output live
 - Auto-cleanup when agents complete
 
-Customize agent models, prompts, and permissions in `oh-my-opencode.json`.
+Customize agent models, prompts, and permissions in `oh-my-openagent.json`.
 
 ## Category System
 
@@ -108,12 +108,12 @@ By combining these two concepts, you can generate optimal agents through `task`.
 | Category             | Default Model                   | Use Cases                                                                                                                   |
 | -------------------- | ------------------------------- | --------------------------------------------------------------------------------------------------------------------------- |
 | `visual-engineering` | `google/gemini-3.1-pro`         | Frontend, UI/UX, design, styling, animation                                                                                 |
-| `ultrabrain`         | `openai/gpt-5.3-codex` (xhigh)  | Deep logical reasoning, complex architecture decisions requiring extensive analysis                                         |
+| `ultrabrain`         | `openai/gpt-5.4` (xhigh)        | Deep logical reasoning, complex architecture decisions requiring extensive analysis                                         |
 | `deep`               | `openai/gpt-5.3-codex` (medium) | Goal-oriented autonomous problem-solving. Thorough research before action. For hairy problems requiring deep understanding. |
 | `artistry`           | `google/gemini-3.1-pro` (high)  | Highly creative/artistic tasks, novel ideas                                                                                 |
-| `quick`              | `anthropic/claude-haiku-4-5`    | Trivial tasks - single file changes, typo fixes, simple modifications                                                       |
+| `quick`              | `openai/gpt-5.4-mini`           | Trivial tasks - single file changes, typo fixes, simple modifications                                                       |
 | `unspecified-low`    | `anthropic/claude-sonnet-4-6`   | Tasks that don't fit other categories, low effort required                                                                  |
-| `unspecified-high`   | `openai/gpt-5.4` (high)         | Tasks that don't fit other categories, high effort required                                                                 |
+| `unspecified-high`   | `anthropic/claude-opus-4-6` (max) | Tasks that don't fit other categories, high effort required                                                               |
 | `writing`            | `google/gemini-3-flash`         | Documentation, prose, technical writing                                                                                     |
 
 ### Usage
@@ -129,7 +129,7 @@ task({
 
 ### Custom Categories
 
-You can define custom categories in `oh-my-opencode.json`.
+You can define custom categories in `oh-my-openagent.json`.
 
 #### Category Configuration Schema
 
@@ -237,7 +237,7 @@ Skills provide specialized workflows with embedded MCP servers and detailed inst
 
 ### Browser Automation Options
 
-Oh-My-OpenCode provides two browser automation providers, configurable via `browser_automation_engine.provider`.
+Oh-My-OpenAgent provides two browser automation providers, configurable via `browser_automation_engine.provider`.
 
 #### Option 1: Playwright MCP (Default)
 
@@ -332,7 +332,7 @@ You can create powerful specialized agents by combining Categories and Skills.
 
 - **Category**: `ultrabrain`
 - **load_skills**: `[]` (pure reasoning)
-- **Effect**: Leverages GPT-5.3 Codex's logical reasoning for in-depth system architecture analysis.
+- **Effect**: Leverages GPT-5.4 xhigh reasoning for in-depth system architecture analysis.
 
 #### The Maintainer (Quick Fixes)
 
@@ -558,7 +558,7 @@ Requires `experimental.task_system: true` in config.
 
 #### Task System Details
 
-**Note on Claude Code Alignment**: This implementation follows Claude Code's internal Task tool signatures (`TaskCreate`, `TaskUpdate`, `TaskList`, `TaskGet`) and field naming conventions (`subject`, `blockedBy`, `blocks`, etc.). However, Anthropic has not published official documentation for these tools. This is Oh My OpenCode's own implementation based on observed Claude Code behavior and internal specifications.
+**Note on Claude Code Alignment**: This implementation follows Claude Code's internal Task tool signatures (`TaskCreate`, `TaskUpdate`, `TaskList`, `TaskGet`) and field naming conventions (`subject`, `blockedBy`, `blocks`, etc.). However, Anthropic has not published official documentation for these tools. This is Oh My OpenAgent's own implementation based on observed Claude Code behavior and internal specifications.
 
 **Task Schema**:
 
@@ -785,7 +785,7 @@ Disable specific hooks in config:
 
 ```json
 {
-  "disabled_hooks": ["comment-checker", "auto-update-checker"]
+  "disabled_hooks": ["comment-checker"]
 }
 ```
 
@@ -844,7 +844,7 @@ When a skill MCP has `oauth` configured:
 Pre-authenticate via CLI:
 
 ```bash
-bunx oh-my-opencode mcp oauth login <server-name> --server-url https://api.example.com
+bunx oh-my-openagent mcp oauth login <server-name> --server-url https://api.example.com
 ```
 
 ## Context Injection

docs/superpowers/plans/2026-03-17-model-settings-compatibility-resolver.md

@@ -0,0 +1,86 @@
+# Model Settings Compatibility Resolver Implementation Plan
+
+> **For agentic workers:** REQUIRED: Use superpowers:subagent-driven-development (if subagents available) or superpowers:executing-plans to implement this plan. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** Centralize compatibility handling for `variant` and `reasoningEffort` so an already-selected model receives the best valid settings for that exact model.
+
+**Architecture:** Introduce a pure shared resolver in `src/shared/` that computes compatible settings and records downgrades/removals. Integrate it first in `chat.params`, then keep Claude-specific effort logic as a thin layer rather than a special-case policy owner.
+
+**Tech Stack:** TypeScript, Bun test, existing shared model normalization/utilities, OpenCode plugin `chat.params` path.
+
+---
+
+### Task 1: Create the pure compatibility resolver
+
+**Files:**
+- Create: `src/shared/model-settings-compatibility.ts`
+- Create: `src/shared/model-settings-compatibility.test.ts`
+- Modify: `src/shared/index.ts`
+
+- [ ] **Step 1: Write failing tests for exact keep behavior**
+- [ ] **Step 2: Write failing tests for downgrade behavior (`max` -> `high`, `xhigh` -> `high` where needed)**
+- [ ] **Step 3: Write failing tests for unsupported-value removal**
+- [ ] **Step 4: Write failing tests for model-family distinctions (Opus vs Sonnet/Haiku, GPT-family variants)**
+- [ ] **Step 5: Implement the pure resolver with explicit capability ladders**
+- [ ] **Step 6: Export the resolver from `src/shared/index.ts`**
+- [ ] **Step 7: Run `bun test src/shared/model-settings-compatibility.test.ts`**
+- [ ] **Step 8: Commit**
+
+### Task 2: Integrate resolver into chat.params
+
+**Files:**
+- Modify: `src/plugin/chat-params.ts`
+- Modify: `src/plugin/chat-params.test.ts`
+
+- [ ] **Step 1: Write failing tests showing `chat.params` applies resolver output to runtime settings**
+- [ ] **Step 2: Ensure tests cover both `variant` and `reasoningEffort` decisions**
+- [ ] **Step 3: Update `chat-params.ts` to call the shared resolver before hook-specific adjustments**
+- [ ] **Step 4: Preserve existing prompt-param-store merging behavior**
+- [ ] **Step 5: Run `bun test src/plugin/chat-params.test.ts`**
+- [ ] **Step 6: Commit**
+
+### Task 3: Re-scope anthropic-effort around the resolver
+
+**Files:**
+- Modify: `src/hooks/anthropic-effort/hook.ts`
+- Modify: `src/hooks/anthropic-effort/index.test.ts`
+
+- [ ] **Step 1: Write failing tests that codify the intended remaining Anthropic-specific behavior after centralization**
+- [ ] **Step 2: Reduce `anthropic-effort` to Claude/Anthropic-specific effort injection where still needed**
+- [ ] **Step 3: Remove duplicated compatibility policy from the hook if the shared resolver now owns it**
+- [ ] **Step 4: Run `bun test src/hooks/anthropic-effort/index.test.ts`**
+- [ ] **Step 5: Commit**
+
+### Task 4: Add integration/regression coverage across real request paths
+
+**Files:**
+- Modify: `src/plugin/chat-params.test.ts`
+- Modify: `src/hooks/anthropic-effort/index.test.ts`
+- Add tests only where needed in nearby suites
+
+- [ ] **Step 1: Add regression test for non-Opus Claude with `variant=max` resolving to compatible settings without ad hoc path-only logic**
+- [ ] **Step 2: Add regression test for GPT-style `reasoningEffort` compatibility**
+- [ ] **Step 3: Add regression test showing supported values remain unchanged**
+- [ ] **Step 4: Run the focused test set**
+- [ ] **Step 5: Commit**
+
+### Task 5: Verify full quality bar
+
+**Files:**
+- No intended code changes
+
+- [ ] **Step 1: Run `bun run typecheck`**
+- [ ] **Step 2: Run a focused suite for the touched files**
+- [ ] **Step 3: If clean, run `bun test`**
+- [ ] **Step 4: Review diff for accidental scope creep**
+- [ ] **Step 5: Commit any final cleanup**
+
+### Task 6: Prepare PR metadata
+
+**Files:**
+- No repo file change required unless docs are updated further
+
+- [ ] **Step 1: Write a human summary explaining this is settings compatibility, not model fallback**
+- [ ] **Step 2: Document scope: Phase 1 covers `variant` and `reasoningEffort` only**
+- [ ] **Step 3: Document explicit non-goals: no model switching, no automatic upscaling in Phase 1**
+- [ ] **Step 4: Request review**

docs/superpowers/specs/2026-03-17-model-settings-compatibility-design.md

@@ -0,0 +1,164 @@
+# Model Settings Compatibility Resolver Design
+
+## Goal
+
+Introduce a central resolver that takes an already-selected model and a set of desired model settings, then returns the best compatible configuration for that exact model.
+
+This is explicitly separate from model fallback.
+
+## Problem
+
+Today, logic for `variant` and `reasoningEffort` compatibility is scattered across multiple places:
+- `hooks/anthropic-effort`
+- `plugin/chat-params`
+- agent/category/fallback config layers
+- delegate/background prompt plumbing
+
+That creates inconsistent behavior:
+- some paths clamp unsupported levels
+- some paths pass them through unchanged
+- some paths silently drop them
+- some paths use model-family-specific assumptions that do not generalize
+
+The result is brittle request behavior even when the chosen model itself is valid.
+
+## Scope
+
+Phase 1 covers only:
+- `variant`
+- `reasoningEffort`
+
+Out of scope for Phase 1:
+- model fallback itself
+- `thinking`
+- `maxTokens`
+- `temperature`
+- `top_p`
+- automatic upward remapping of settings
+
+## Desired behavior
+
+Given a fixed model and desired settings:
+1. If a desired value is supported, keep it.
+2. If not supported, downgrade to the nearest lower compatible value.
+3. If no compatible value exists, drop the field.
+4. Do not switch models.
+5. Do not automatically upgrade settings in Phase 1.
+
+## Architecture
+
+Add a central module:
+- `src/shared/model-settings-compatibility.ts`
+
+Core API:
+
+```ts
+type DesiredModelSettings = {
+  variant?: string
+  reasoningEffort?: string
+}
+
+type ModelSettingsCompatibilityInput = {
+  providerID: string
+  modelID: string
+  desired: DesiredModelSettings
+}
+
+type ModelSettingsCompatibilityChange = {
+  field: "variant" | "reasoningEffort"
+  from: string
+  to?: string
+  reason: string
+}
+
+type ModelSettingsCompatibilityResult = {
+  variant?: string
+  reasoningEffort?: string
+  changes: ModelSettingsCompatibilityChange[]
+}
+```
+
+## Compatibility model
+
+Phase 1 should be **metadata-first where the platform exposes reliable capability data**, and only fall back to family-based rules when that metadata is absent.
+
+### Variant compatibility
+
+Preferred source of truth:
+- OpenCode/provider model metadata (`variants`)
+
+Fallback when metadata is unavailable:
+- family-based ladders
+
+Examples of fallback ladders:
+- Claude Opus family: `low`, `medium`, `high`, `max`
+- Claude Sonnet/Haiku family: `low`, `medium`, `high`
+- OpenAI GPT family: conservative family fallback only when metadata is missing
+- Unknown family: drop unsupported values conservatively
+
+### Reasoning effort compatibility
+
+Current Phase 1 source of truth:
+- conservative model/provider family heuristics
+
+Reason:
+- the currently available OpenCode SDK/provider metadata exposes model `variants`, but does not expose an equivalent per-model capability list for `reasoningEffort` levels
+
+Examples:
+- GPT/OpenAI-style models: `low`, `medium`, `high`, `xhigh` where supported by family heuristics
+- Claude family via current OpenCode path: treat `reasoningEffort` as unsupported in Phase 1 and remove it
+
+The resolver should remain pure model/settings logic only. Transport restrictions remain the responsibility of the request-building path.
+
+## Separation of concerns
+
+This design intentionally separates:
+- model selection (`resolveModel...`, fallback chains)
+- settings compatibility (this resolver)
+- request transport compatibility (`chat.params`, prompt body constraints)
+
+That keeps responsibilities clear:
+- choose model first
+- normalize settings second
+- build request third
+
+## First integration point
+
+Phase 1 should first integrate into `chat.params`.
+
+Why:
+- it is already the centralized path for request-time tuning
+- it can influence provider-facing options without leaking unsupported fields into prompt payload bodies
+- it avoids trying to patch every prompt constructor at once
+
+## Rollout plan
+
+### Phase 1
+- add resolver module and tests
+- integrate into `chat.params`
+- migrate `anthropic-effort` to either use the resolver or become a thin Claude-specific supplement around it
+
+### Phase 2
+- expand to `thinking`, `maxTokens`, `temperature`, `top_p`
+- formalize request-path capability tables if needed
+
+### Phase 3
+- centralize all variant/reasoning normalization away from scattered hooks and ad hoc callers
+
+## Risks
+
+- Overfitting family rules to current model naming conventions
+- Accidentally changing request semantics on paths that currently rely on implicit behavior
+- Mixing provider transport limitations with model capability logic
+
+## Mitigations
+
+- Keep resolver pure and narrowly scoped in Phase 1
+- Add explicit regression tests for keep/downgrade/drop decisions
+- Integrate at one central point first (`chat.params`)
+- Preserve existing behavior where desired values are already valid
+
+## Recommendation
+
+Proceed with the central resolver as a new, isolated implementation in a dedicated branch/worktree.
+This is the clean long-term path and is more reviewable than continuing to add special-case clamps in hooks.

docs/troubleshooting/ollama.md

@@ -4,7 +4,7 @@
 
 ### Problem
 
-When using Ollama as a provider with oh-my-opencode agents, you may encounter:
+When using Ollama as a provider with oh-my-openagent agents, you may encounter:
 
 ```
 JSON Parse error: Unexpected EOF
@@ -26,7 +26,7 @@ Claude Code SDK expects a single JSON object, not multiple NDJSON lines, causing
 **Why this happens:**
 - **Ollama API**: Returns streaming responses as NDJSON by design
 - **Claude Code SDK**: Doesn't properly handle NDJSON responses for tool calls
-- **oh-my-opencode**: Passes through the SDK's behavior (can't fix at this layer)
+- **oh-my-openagent**: Passes through the SDK's behavior (can't fix at this layer)
 
 ## Solutions
 
@@ -114,7 +114,7 @@ curl -s http://localhost:11434/api/chat \
 
 ## Related Issues
 
-- **oh-my-opencode**: https://github.com/code-yeongyu/oh-my-openagent/issues/1124
+- **oh-my-openagent**: https://github.com/code-yeongyu/oh-my-openagent/issues/1124
 - **Ollama API Docs**: https://github.com/ollama/ollama/blob/main/docs/api.md
 
 ## Getting Help