release: v3.9.0

The auto-commit section instructed Hephaestus to automatically commit after
implementation work. Users who didn't know about this behavior would get
surprise commits — a trust-breaking behavioral change flagged by 5 Oracle
reviews as the sole publish blocker for 3.9.0.

.github/workflows/publish-platform.yml

@@ -35,15 +35,15 @@ jobs:
   # - Uploads compressed artifacts for the publish job
   # =============================================================================
   build:
-    runs-on: ${{ matrix.platform == 'windows-x64' && 'windows-latest' || 'ubuntu-latest' }}
+    runs-on: ${{ startsWith(matrix.platform, 'windows-') && 'windows-latest' || 'ubuntu-latest' }}
     defaults:
       run:
         shell: bash
     strategy:
       fail-fast: false
-      max-parallel: 7
+      max-parallel: 11
       matrix:
-        platform: [darwin-arm64, darwin-x64, linux-x64, linux-arm64, linux-x64-musl, linux-arm64-musl, windows-x64]
+        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:
       - uses: actions/checkout@v4
 
@@ -95,14 +95,18 @@ jobs:
             case "$PLATFORM" in
               darwin-arm64) TARGET="bun-darwin-arm64" ;;
               darwin-x64) TARGET="bun-darwin-x64" ;;
+              darwin-x64-baseline) TARGET="bun-darwin-x64-baseline" ;;
               linux-x64) TARGET="bun-linux-x64" ;;
+              linux-x64-baseline) TARGET="bun-linux-x64-baseline" ;;
               linux-arm64) TARGET="bun-linux-arm64" ;;
               linux-x64-musl) TARGET="bun-linux-x64-musl" ;;
+              linux-x64-musl-baseline) TARGET="bun-linux-x64-musl-baseline" ;;
               linux-arm64-musl) TARGET="bun-linux-arm64-musl" ;;
               windows-x64) TARGET="bun-windows-x64" ;;
+              windows-x64-baseline) TARGET="bun-windows-x64-baseline" ;;
             esac
             
-            if [ "$PLATFORM" = "windows-x64" ]; then
+            if [[ "$PLATFORM" == windows-* ]]; then
               OUTPUT="packages/${PLATFORM}/bin/oh-my-opencode.exe"
             else
               OUTPUT="packages/${PLATFORM}/bin/oh-my-opencode"
@@ -119,7 +123,7 @@ jobs:
           PLATFORM="${{ matrix.platform }}"
           cd packages/${PLATFORM}
           
-          if [ "$PLATFORM" = "windows-x64" ]; then
+          if [[ "$PLATFORM" == windows-* ]]; then
             # Windows: use 7z (pre-installed on windows-latest)
             7z a -tzip ../../binary-${PLATFORM}.zip bin/ package.json
           else
@@ -155,7 +159,7 @@ jobs:
       fail-fast: false
       max-parallel: 2
       matrix:
-        platform: [darwin-arm64, darwin-x64, linux-x64, linux-arm64, linux-x64-musl, linux-arm64-musl, windows-x64]
+        platform: [darwin-arm64, darwin-x64, darwin-x64-baseline, linux-x64, linux-x64-baseline, linux-arm64, linux-x64-musl, linux-x64-musl-baseline, linux-arm64-musl, windows-x64, windows-x64-baseline]
     steps:
       - name: Check if already published
         id: check
@@ -184,7 +188,7 @@ jobs:
           PLATFORM="${{ matrix.platform }}"
           mkdir -p packages/${PLATFORM}
           
-          if [ "$PLATFORM" = "windows-x64" ]; then
+          if [[ "$PLATFORM" == windows-* ]]; then
             unzip binary-${PLATFORM}.zip -d packages/${PLATFORM}/
           else
             tar -xzvf binary-${PLATFORM}.tar.gz -C packages/${PLATFORM}/

.github/workflows/publish.yml

@@ -189,7 +189,7 @@ jobs:
           VERSION="${{ steps.version.outputs.version }}"
           jq --arg v "$VERSION" '.version = $v' package.json > tmp.json && mv tmp.json package.json
           
-          for platform in darwin-arm64 darwin-x64 linux-x64 linux-arm64 linux-x64-musl linux-arm64-musl windows-x64; do
+          for platform in 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; do
             jq --arg v "$VERSION" '.version = $v' "packages/${platform}/package.json" > tmp.json
             mv tmp.json "packages/${platform}/package.json"
           done

assets/oh-my-opencode.schema.json

@@ -24,19 +24,7 @@
     "disabled_agents": {
       "type": "array",
       "items": {
-        "type": "string",
-        "enum": [
-          "sisyphus",
-          "hephaestus",
-          "prometheus",
-          "oracle",
-          "librarian",
-          "explore",
-          "multimodal-looker",
-          "metis",
-          "momus",
-          "atlas"
-        ]
+        "type": "string"
       }
     },
     "disabled_skills": {

benchmarks/bun.lock

@@ -0,0 +1,62 @@
+{
+  "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/headless.ts

@@ -0,0 +1,193 @@
+#!/usr/bin/env bun
+import { readFile, writeFile, mkdir } from "node:fs/promises"
+import { join, dirname } from "node:path"
+import { stepCountIs, streamText, type CoreMessage } from "ai"
+import { tool } from "ai"
+import { createFriendli } from "@friendliai/ai-provider"
+import { z } from "zod"
+import { formatHashLines } from "../src/tools/hashline-edit/hash-computation"
+import { normalizeHashlineEdits } from "../src/tools/hashline-edit/normalize-edits"
+import { applyHashlineEditsWithReport } from "../src/tools/hashline-edit/edit-operations"
+import { canonicalizeFileText, restoreFileText } from "../src/tools/hashline-edit/file-text-canonicalization"
+
+const DEFAULT_MODEL = "MiniMaxAI/MiniMax-M2.5"
+const MAX_STEPS = 50
+const sessionId = `bench-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`
+
+const emit = (event: Record<string, unknown>) =>
+  console.log(JSON.stringify({ sessionId, timestamp: new Date().toISOString(), ...event }))
+
+// ── CLI ──────────────────────────────────────────────────────
+function parseArgs(): { prompt: string; modelId: string } {
+  const args = process.argv.slice(2)
+  let prompt = ""
+  let modelId = DEFAULT_MODEL
+  for (let i = 0; i < args.length; i++) {
+    if ((args[i] === "-p" || args[i] === "--prompt") && args[i + 1]) {
+      prompt = args[++i]
+    } else if ((args[i] === "-m" || args[i] === "--model") && args[i + 1]) {
+      modelId = args[++i]
+    } else if (args[i] === "--reasoning-mode" && args[i + 1]) {
+      i++ // consume
+    }
+    // --no-translate, --think consumed silently
+  }
+  if (!prompt) {
+    console.error("Usage: bun run benchmarks/headless.ts -p <prompt> [-m <model>]")
+    process.exit(1)
+  }
+  return { prompt, modelId }
+}
+
+// ── Tools ────────────────────────────────────────────────────
+const readFileTool = tool({
+  description: "Read a file with hashline-tagged content (LINE#ID format)",
+  inputSchema: z.object({ path: z.string().describe("File path") }),
+  execute: async ({ path }) => {
+    const fullPath = join(process.cwd(), path)
+    try {
+      const content = await readFile(fullPath, "utf-8")
+      const lines = content.split("\n")
+      const tagged = formatHashLines(content)
+      return `OK - read file\npath: ${path}\nlines: ${lines.length}\n\n${tagged}`
+    } catch {
+      return `Error: File not found: ${path}`
+    }
+  },
+})
+
+const editFileTool = tool({
+  description: "Edit a file using hashline anchors (LINE#ID format)",
+  inputSchema: z.object({
+    path: z.string(),
+    edits: z.array(
+      z.object({
+        op: z.enum(["replace", "append", "prepend"]),
+        pos: z.string().optional(),
+        end: z.string().optional(),
+        lines: z.union([z.array(z.string()), z.string(), z.null()]),
+      })
+    ).min(1),
+  }),
+  execute: async ({ path, edits }) => {
+    const fullPath = join(process.cwd(), path)
+    try {
+      let rawContent = ""
+      let exists = true
+      try {
+        rawContent = await readFile(fullPath, "utf-8")
+      } catch {
+        exists = false
+      }
+
+      const normalized = normalizeHashlineEdits(edits)
+
+      if (!exists) {
+        const canCreate = normalized.every(
+          (e) => (e.op === "append" || e.op === "prepend") && !e.pos
+        )
+        if (!canCreate) return `Error: File not found: ${path}`
+      }
+
+      const envelope = canonicalizeFileText(rawContent)
+      const result = applyHashlineEditsWithReport(envelope.content, normalized)
+
+      if (result.content === envelope.content) {
+        return `Error: No changes made to ${path}. The edits produced identical content.`
+      }
+
+      const writeContent = restoreFileText(result.content, envelope)
+      await mkdir(dirname(fullPath), { recursive: true })
+      await writeFile(fullPath, writeContent, "utf-8")
+
+      const oldLineCount = rawContent.split("\n").length
+      const newLineCount = writeContent.split("\n").length
+      const delta = newLineCount - oldLineCount
+      const sign = delta > 0 ? "+" : ""
+      const action = exists ? "Updated" : "Created"
+      return `${action} ${path}\n${edits.length} edit(s) applied, ${sign}${delta} line(s)`
+    } catch (error) {
+      return `Error: ${error instanceof Error ? error.message : String(error)}`
+    }
+  },
+})
+
+// ── Agent Loop ───────────────────────────────────────────────
+async function run() {
+  const { prompt, modelId } = parseArgs()
+
+  const friendli = createFriendli({ apiKey: process.env.FRIENDLI_TOKEN! })
+  const model = friendli(modelId)
+  const tools = { read_file: readFileTool, edit_file: editFileTool }
+
+  emit({ type: "user", content: prompt })
+
+  const messages: CoreMessage[] = [{ role: "user", content: prompt }]
+  const system =
+    "You are a code editing assistant. Use read_file to read files and edit_file to edit them. " +
+    "Always read a file before editing it to get fresh LINE#ID anchors."
+
+  for (let step = 0; step < MAX_STEPS; step++) {
+    const stream = streamText({
+      model,
+      tools,
+      messages,
+      system,
+      stopWhen: stepCountIs(1),
+    })
+
+    let currentText = ""
+    for await (const part of stream.fullStream) {
+      switch (part.type) {
+        case "text-delta":
+          currentText += part.text
+          break
+        case "tool-call":
+          emit({
+            type: "tool_call",
+            tool_call_id: part.toolCallId,
+            tool_name: part.toolName,
+            tool_input: part.args,
+            model: modelId,
+          })
+          break
+        case "tool-result": {
+          const output = typeof part.result === "string" ? part.result : JSON.stringify(part.result)
+          const isError = typeof output === "string" && output.startsWith("Error:")
+          emit({
+            type: "tool_result",
+            tool_call_id: part.toolCallId,
+            output,
+            ...(isError ? { error: output } : {}),
+          })
+          break
+      }
+    }
+
+    const response = await stream.response
+    messages.push(...response.messages)
+
+    const finishReason = await stream.finishReason
+    if (finishReason !== "tool-calls") {
+      if (currentText.trim()) {
+        emit({ type: "assistant", content: currentText, model: modelId })
+      }
+      break
+    }
+  }
+}
+
+// ── Signal + Startup ─────────────────────────────────────────
+process.once("SIGINT", () => process.exit(0))
+process.once("SIGTERM", () => process.exit(143))
+
+const startTime = Date.now()
+run()
+  .catch((error) => {
+    emit({ type: "error", error: error instanceof Error ? error.message : String(error) })
+    process.exit(1)
+  })
+  .then(() => {
+    const elapsed = ((Date.now() - startTime) / 1000).toFixed(2)
+    console.error(`[headless] Completed in ${elapsed}s`)
+  })

benchmarks/package.json

@@ -0,0 +1,19 @@
+{
+  "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"
+  }
+}

benchmarks/test-edit-ops.ts

@@ -0,0 +1,808 @@
+#!/usr/bin/env bun
+/**
+ * Comprehensive headless edit_file stress test: 21 operation types
+ *
+ * Tests: 5 basic ops + 10 creative cases + 6 whitespace cases
+ * Each runs via headless mode with its own demo file + prompt.
+ *
+ * Usage:
+ *   bun run scripts/test-headless-edit-ops.ts [-m <model>] [--provider <provider>]
+ */
+
+import { spawn } from "node:child_process";
+import { mkdirSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join, resolve } from "node:path";
+
+// ── CLI arg passthrough ───────────────────────────────────────
+const extraArgs: string[] = [];
+const rawArgs = process.argv.slice(2);
+for (let i = 0; i < rawArgs.length; i++) {
+  const arg = rawArgs[i];
+  if (
+    (arg === "-m" || arg === "--model" || arg === "--provider") &&
+    i + 1 < rawArgs.length
+  ) {
+    extraArgs.push(arg, rawArgs[i + 1]);
+    i++;
+  } else if (arg === "--think" || arg === "--no-translate") {
+    extraArgs.push(arg);
+  } else if (arg === "--reasoning-mode" && i + 1 < rawArgs.length) {
+    extraArgs.push(arg, rawArgs[i + 1]);
+    i++;
+  }
+}
+
+// ── Colors ────────────────────────────────────────────────────
+const BOLD = "\x1b[1m";
+const GREEN = "\x1b[32m";
+const RED = "\x1b[31m";
+const YELLOW = "\x1b[33m";
+const DIM = "\x1b[2m";
+const CYAN = "\x1b[36m";
+const RESET = "\x1b[0m";
+
+const pass = (msg: string) => console.log(`  ${GREEN}✓${RESET} ${msg}`);
+const fail = (msg: string) => console.log(`  ${RED}✗${RESET} ${msg}`);
+const info = (msg: string) => console.log(`  ${DIM}${msg}${RESET}`);
+const warn = (msg: string) => console.log(`  ${YELLOW}⚠${RESET} ${msg}`);
+
+// ── Test case definition ─────────────────────────────────────
+interface TestCase {
+  fileContent: string;
+  fileName: string;
+  name: string;
+  prompt: string;
+  validate: (content: string) => { passed: boolean; reason: string };
+}
+
+const TEST_CASES: TestCase[] = [
+  {
+    name: "1. Replace single line",
+    fileName: "config.txt",
+    fileContent: [
+      "host: localhost",
+      "port: 3000",
+      "debug: false",
+      "timeout: 30",
+      "retries: 3",
+    ].join("\n"),
+    prompt: [
+      "Follow these steps exactly:",
+      "Step 1: Call read_file on config.txt.",
+      "Step 2: Note the anchor for the port line (line 2).",
+      "Step 3: Call edit_file with path='config.txt' and edits containing ONE object:",
+      "  { op: 'replace', pos: '<line2 anchor>', lines: ['port: 8080'] }",
+      "IMPORTANT: pos must be ONLY the anchor (like '2#KB'). lines must be a SEPARATE array field with the new content.",
+    ].join(" "),
+    validate: (content) => {
+      const has8080 = content.includes("port: 8080");
+      const has3000 = content.includes("port: 3000");
+      if (has8080 && !has3000) {
+        return { passed: true, reason: "port changed to 8080" };
+      }
+      if (has3000) {
+        return { passed: false, reason: "port still 3000 — edit not applied" };
+      }
+      return {
+        passed: false,
+        reason: `unexpected content: ${content.slice(0, 100)}`,
+      };
+    },
+  },
+  {
+    name: "2. Append after line",
+    fileName: "fruits.txt",
+    fileContent: ["apple", "banana", "cherry"].join("\n"),
+    prompt:
+      "Read fruits.txt with read_file. Then use edit_file with op='append' to insert a new line 'grape' after the 'banana' line. Use pos='LINE#HASH' of the banana line and lines=['grape'].",
+    validate: (content) => {
+      const lines = content.trim().split("\n");
+      const bananaIdx = lines.findIndex((l) => l.trim() === "banana");
+      const grapeIdx = lines.findIndex((l) => l.trim() === "grape");
+      if (grapeIdx === -1) {
+        return { passed: false, reason: '"grape" not found in file' };
+      }
+      if (bananaIdx === -1) {
+        return { passed: false, reason: '"banana" was removed' };
+      }
+      if (grapeIdx !== bananaIdx + 1) {
+        return {
+          passed: false,
+          reason: `"grape" at line ${grapeIdx + 1} but expected after "banana" at line ${bananaIdx + 1}`,
+        };
+      }
+      if (lines.length !== 4) {
+        return {
+          passed: false,
+          reason: `expected 4 lines, got ${lines.length}`,
+        };
+      }
+      return {
+        passed: true,
+        reason: '"grape" correctly appended after "banana"',
+      };
+    },
+  },
+  {
+    name: "3. Prepend before line",
+    fileName: "code.txt",
+    fileContent: ["function greet() {", '  return "hello";', "}"].join("\n"),
+    prompt:
+      "Read code.txt with read_file. Then use edit_file with op='prepend' to add '// Greeting function' before the function line. Use pos='LINE#HASH' of the function line and lines=['// Greeting function'].",
+    validate: (content) => {
+      const lines = content.trim().split("\n");
+      const commentIdx = lines.findIndex(
+        (l) => l.trim().startsWith("//") && l.toLowerCase().includes("greet")
+      );
+      const funcIdx = lines.findIndex((l) =>
+        l.trim().startsWith("function greet")
+      );
+      if (commentIdx === -1) {
+        return { passed: false, reason: "comment line not found" };
+      }
+      if (funcIdx === -1) {
+        return { passed: false, reason: '"function greet" line was removed' };
+      }
+      if (commentIdx !== funcIdx - 1) {
+        return {
+          passed: false,
+          reason: `comment at line ${commentIdx + 1} but function at ${funcIdx + 1} — not directly before`,
+        };
+      }
+      return {
+        passed: true,
+        reason: "comment correctly prepended before function",
+      };
+    },
+  },
+  {
+    name: "4. Range replace (multi-line → single line)",
+    fileName: "log.txt",
+    fileContent: [
+      "=== Log Start ===",
+      "INFO: started",
+      "WARN: slow query",
+      "ERROR: timeout",
+      "INFO: recovered",
+      "=== Log End ===",
+    ].join("\n"),
+    prompt: [
+      "Follow these steps exactly:",
+      "Step 1: Call read_file on log.txt to see line anchors.",
+      "Step 2: Note the anchor for 'WARN: slow query' (line 3) and 'ERROR: timeout' (line 4).",
+      "Step 3: Call edit_file with path='log.txt' and edits containing ONE object with THREE separate JSON fields:",
+      "  { op: 'replace', pos: '<line3 anchor>', end: '<line4 anchor>', lines: ['RESOLVED: issues cleared'] }",
+      "CRITICAL: pos, end, and lines are THREE SEPARATE JSON fields. pos is ONLY '3#XX'. end is ONLY '4#YY'. lines is ['RESOLVED: issues cleared'].",
+      "If edit_file fails or errors, use write_file to write the complete correct file content instead.",
+      "The correct final content should be: === Log Start ===, INFO: started, RESOLVED: issues cleared, INFO: recovered, === Log End ===",
+      "Do not make any other changes.",
+    ].join(" "),
+    validate: (content) => {
+      const lines = content.trim().split("\n");
+      const hasResolved = lines.some(
+        (l) => l.trim() === "RESOLVED: issues cleared"
+      );
+      const hasWarn = content.includes("WARN: slow query");
+      const hasError = content.includes("ERROR: timeout");
+      if (!hasResolved) {
+        return {
+          passed: false,
+          reason: '"RESOLVED: issues cleared" not found',
+        };
+      }
+      if (hasWarn || hasError) {
+        return { passed: false, reason: "old WARN/ERROR lines still present" };
+      }
+      // Core assertion: 2 old lines removed, 1 new line added = net -1 line
+      // Allow slight overshoot from model adding extra content
+      if (lines.length < 4 || lines.length > 6) {
+        return {
+          passed: false,
+          reason: `expected ~5 lines, got ${lines.length}`,
+        };
+      }
+      return {
+        passed: true,
+        reason: "range replace succeeded — 2 lines → 1 line",
+      };
+    },
+  },
+  {
+    name: "5. Delete line",
+    fileName: "settings.txt",
+    fileContent: [
+      "mode: production",
+      "debug: true",
+      "cache: enabled",
+      "log_level: info",
+    ].join("\n"),
+    prompt: [
+      "Follow these steps exactly:",
+      "Step 1: Call read_file on settings.txt to see line anchors.",
+      "Step 2: Note the anchor for 'debug: true' (line 2).",
+      "Step 3: Call edit_file with path='settings.txt' and edits containing ONE object:",
+      "  { op: 'replace', pos: '<line2 anchor>', lines: [] }",
+      "IMPORTANT: lines must be an empty array [] to delete the line. pos must be ONLY the anchor like '2#SR'.",
+    ].join(" "),
+    validate: (content) => {
+      const lines = content.trim().split("\n");
+      const hasDebug = content.includes("debug: true");
+      if (hasDebug) {
+        return { passed: false, reason: '"debug: true" still present' };
+      }
+      if (lines.length !== 3) {
+        return {
+          passed: false,
+          reason: `expected 3 lines, got ${lines.length}`,
+        };
+      }
+      if (
+        !(
+          content.includes("mode: production") &&
+          content.includes("cache: enabled")
+        )
+      ) {
+        return { passed: false, reason: "other lines were removed" };
+      }
+      return { passed: true, reason: '"debug: true" successfully deleted' };
+    },
+  },
+
+  // ── Creative cases (6-15) ────────────────────────────────────
+  {
+    name: "6. Batch edit — two replacements in one call",
+    fileName: "batch.txt",
+    fileContent: ["red", "green", "blue", "yellow"].join("\n"),
+    prompt: [
+      "Read batch.txt with read_file.",
+      "Then call edit_file ONCE with path='batch.txt' and edits containing TWO objects:",
+      "  1) { op: 'replace', pos: '<line1 anchor>', lines: ['crimson'] }",
+      "  2) { op: 'replace', pos: '<line3 anchor>', lines: ['navy'] }",
+      "Both edits must be in the SAME edits array in a single edit_file call.",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (!c.includes("crimson")) return { passed: false, reason: "'crimson' not found" };
+      if (!c.includes("navy")) return { passed: false, reason: "'navy' not found" };
+      if (c.includes("red")) return { passed: false, reason: "'red' still present" };
+      if (c.includes("blue")) return { passed: false, reason: "'blue' still present" };
+      if (lines.length !== 4) return { passed: false, reason: `expected 4 lines, got ${lines.length}` };
+      return { passed: true, reason: "both lines replaced in single call" };
+    },
+  },
+  {
+    name: "7. Line expansion — 1 line → 3 lines",
+    fileName: "expand.txt",
+    fileContent: ["header", "TODO: implement", "footer"].join("\n"),
+    prompt: [
+      "Read expand.txt with read_file.",
+      "Replace the 'TODO: implement' line (line 2) with THREE lines:",
+      "  'step 1: init', 'step 2: process', 'step 3: cleanup'",
+      "Use edit_file with op='replace', pos=<line2 anchor>, lines=['step 1: init', 'step 2: process', 'step 3: cleanup'].",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (c.includes("TODO")) return { passed: false, reason: "TODO line still present" };
+      if (!c.includes("step 1: init")) return { passed: false, reason: "'step 1: init' not found" };
+      if (!c.includes("step 3: cleanup")) return { passed: false, reason: "'step 3: cleanup' not found" };
+      if (lines.length !== 5) return { passed: false, reason: `expected 5 lines, got ${lines.length}` };
+      return { passed: true, reason: "1 line expanded to 3 lines" };
+    },
+  },
+  {
+    name: "8. Append at EOF",
+    fileName: "eof.txt",
+    fileContent: ["line one", "line two"].join("\n"),
+    prompt: [
+      "Read eof.txt with read_file.",
+      "Use edit_file to append 'line three' after the LAST line of the file.",
+      "Use op='append', pos=<last line anchor>, lines=['line three'].",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (!c.includes("line three")) return { passed: false, reason: "'line three' not found" };
+      if (lines[lines.length - 1].trim() !== "line three")
+        return { passed: false, reason: "'line three' not at end" };
+      if (lines.length !== 3) return { passed: false, reason: `expected 3 lines, got ${lines.length}` };
+      return { passed: true, reason: "appended at EOF" };
+    },
+  },
+  {
+    name: "9. Special characters in content",
+    fileName: "special.json",
+    fileContent: [
+      '{',
+      '  "name": "old-value",',
+      '  "count": 42',
+      '}',
+    ].join("\n"),
+    prompt: [
+      "Read special.json with read_file.",
+      'Replace the line containing \"name\": \"old-value\" with \"name\": \"new-value\".',
+      "Use edit_file with op='replace', pos=<that line's anchor>, lines=['  \"name\": \"new-value\",'].",
+    ].join(" "),
+    validate: (c) => {
+      if (c.includes("old-value")) return { passed: false, reason: "'old-value' still present" };
+      if (!c.includes('"new-value"')) return { passed: false, reason: "'new-value' not found" };
+      if (!c.includes('"count": 42')) return { passed: false, reason: "other content was modified" };
+      return { passed: true, reason: "JSON value replaced with special chars intact" };
+    },
+  },
+  {
+    name: "10. Replace first line",
+    fileName: "first.txt",
+    fileContent: ["OLD HEADER", "body content", "footer"].join("\n"),
+    prompt: [
+      "Read first.txt with read_file.",
+      "Replace the very first line 'OLD HEADER' with 'NEW HEADER'.",
+      "Use edit_file with op='replace', pos=<line1 anchor>, lines=['NEW HEADER'].",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (c.includes("OLD HEADER")) return { passed: false, reason: "'OLD HEADER' still present" };
+      if (lines[0].trim() !== "NEW HEADER") return { passed: false, reason: "first line is not 'NEW HEADER'" };
+      if (!c.includes("body content")) return { passed: false, reason: "body was modified" };
+      return { passed: true, reason: "first line replaced" };
+    },
+  },
+  {
+    name: "11. Replace last line",
+    fileName: "last.txt",
+    fileContent: ["alpha", "bravo", "OLD_FOOTER"].join("\n"),
+    prompt: [
+      "Read last.txt with read_file.",
+      "Replace the last line 'OLD_FOOTER' with 'NEW_FOOTER'.",
+      "Use edit_file with op='replace', pos=<last line anchor>, lines=['NEW_FOOTER'].",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (c.includes("OLD_FOOTER")) return { passed: false, reason: "'OLD_FOOTER' still present" };
+      if (lines[lines.length - 1].trim() !== "NEW_FOOTER")
+        return { passed: false, reason: "last line is not 'NEW_FOOTER'" };
+      return { passed: true, reason: "last line replaced" };
+    },
+  },
+  {
+    name: "12. Adjacent line edits",
+    fileName: "adjacent.txt",
+    fileContent: ["aaa", "bbb", "ccc", "ddd"].join("\n"),
+    prompt: [
+      "Read adjacent.txt with read_file.",
+      "Replace line 2 ('bbb') with 'BBB' and line 3 ('ccc') with 'CCC'.",
+      "Use edit_file with TWO edits in the same call:",
+      "  { op: 'replace', pos: <line2 anchor>, lines: ['BBB'] }",
+      "  { op: 'replace', pos: <line3 anchor>, lines: ['CCC'] }",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (c.includes("bbb")) return { passed: false, reason: "'bbb' still present" };
+      if (c.includes("ccc")) return { passed: false, reason: "'ccc' still present" };
+      if (!c.includes("BBB")) return { passed: false, reason: "'BBB' not found" };
+      if (!c.includes("CCC")) return { passed: false, reason: "'CCC' not found" };
+      if (lines.length !== 4) return { passed: false, reason: `expected 4 lines, got ${lines.length}` };
+      return { passed: true, reason: "two adjacent lines replaced" };
+    },
+  },
+  {
+    name: "13. Prepend multi-line block",
+    fileName: "block.py",
+    fileContent: ["def main():", "    print('hello')", "", "main()"].join("\n"),
+    prompt: [
+      "Read block.py with read_file.",
+      "Prepend a 2-line comment block before 'def main():' (line 1).",
+      "The two lines are: '# Author: test' and '# Date: 2025-01-01'.",
+      "Use edit_file with op='prepend', pos=<line1 anchor>, lines=['# Author: test', '# Date: 2025-01-01'].",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (!c.includes("# Author: test")) return { passed: false, reason: "author comment not found" };
+      if (!c.includes("# Date: 2025-01-01")) return { passed: false, reason: "date comment not found" };
+      const defIdx = lines.findIndex((l) => l.startsWith("def main"));
+      const authorIdx = lines.findIndex((l) => l.includes("Author"));
+      if (authorIdx >= defIdx) return { passed: false, reason: "comments not before def" };
+      return { passed: true, reason: "2-line block prepended before function" };
+    },
+  },
+  {
+    name: "14. Delete range — 3 consecutive lines",
+    fileName: "cleanup.txt",
+    fileContent: ["keep1", "remove-a", "remove-b", "remove-c", "keep2"].join("\n"),
+    prompt: [
+      "Read cleanup.txt with read_file.",
+      "Delete lines 2-4 ('remove-a', 'remove-b', 'remove-c') using a single range replace.",
+      "Use edit_file with op='replace', pos=<line2 anchor>, end=<line4 anchor>, lines=[].",
+      "An empty lines array deletes the range.",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (c.includes("remove")) return { passed: false, reason: "'remove' lines still present" };
+      if (!c.includes("keep1")) return { passed: false, reason: "'keep1' was deleted" };
+      if (!c.includes("keep2")) return { passed: false, reason: "'keep2' was deleted" };
+      if (lines.length !== 2) return { passed: false, reason: `expected 2 lines, got ${lines.length}` };
+      return { passed: true, reason: "3 consecutive lines deleted via range" };
+    },
+  },
+  {
+    name: "15. Replace with duplicate-content line",
+    fileName: "dupes.txt",
+    fileContent: ["item", "item", "item", "item"].join("\n"),
+    prompt: [
+      "Read dupes.txt with read_file. All 4 lines have the same text 'item'.",
+      "Replace ONLY line 3 with 'CHANGED'. Do NOT modify any other line.",
+      "Use edit_file with op='replace', pos=<line3 anchor>, lines=['CHANGED'].",
+      "The anchor hash uniquely identifies line 3 even though the content is identical.",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (!c.includes("CHANGED")) return { passed: false, reason: "'CHANGED' not found" };
+      const changedCount = lines.filter((l) => l.trim() === "CHANGED").length;
+      const itemCount = lines.filter((l) => l.trim() === "item").length;
+      if (changedCount !== 1) return { passed: false, reason: `expected 1 CHANGED, got ${changedCount}` };
+      if (itemCount !== 3) return { passed: false, reason: `expected 3 item lines, got ${itemCount}` };
+      if (lines.length !== 4) return { passed: false, reason: `expected 4 lines, got ${lines.length}` };
+      return { passed: true, reason: "only line 3 changed among duplicates" };
+    },
+  },
+
+  // ── Whitespace cases (16-21) ──────────────────────────────────
+  {
+    name: "16. Fix indentation — 2 spaces → 4 spaces",
+    fileName: "indent.js",
+    fileContent: ["function foo() {", "  const x = 1;", "  return x;", "}"].join("\n"),
+    prompt: [
+      "Read indent.js with read_file.",
+      "Replace line 2 '  const x = 1;' (2-space indent) with '    const x = 1;' (4-space indent).",
+      "Use edit_file with op='replace', pos=<line2 anchor>, lines=['    const x = 1;'].",
+      "The ONLY change is the indentation: 2 spaces → 4 spaces. Content stays the same.",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.split("\n");
+      const line2 = lines[1];
+      if (!line2) return { passed: false, reason: "line 2 missing" };
+      if (line2 === "    const x = 1;") return { passed: true, reason: "indentation fixed to 4 spaces" };
+      if (line2 === "  const x = 1;") return { passed: false, reason: "still 2-space indent" };
+      return { passed: false, reason: `unexpected line 2: '${line2}'` };
+    },
+  },
+  {
+    name: "17. Replace preserving leading whitespace",
+    fileName: "preserve.py",
+    fileContent: [
+      "class Foo:",
+      "    def old_method(self):",
+      "        pass",
+    ].join("\n"),
+    prompt: [
+      "Read preserve.py with read_file.",
+      "Replace line 2 '    def old_method(self):' with '    def new_method(self):'.",
+      "Keep the 4-space indentation. Only change the method name.",
+      "Use edit_file with op='replace', pos=<line2 anchor>, lines=['    def new_method(self):'].",
+    ].join(" "),
+    validate: (c) => {
+      if (c.includes("old_method")) return { passed: false, reason: "'old_method' still present" };
+      const lines = c.split("\n");
+      const methodLine = lines.find((l) => l.includes("new_method"));
+      if (!methodLine) return { passed: false, reason: "'new_method' not found" };
+      if (!methodLine.startsWith("    ")) return { passed: false, reason: "indentation lost" };
+      return { passed: true, reason: "method renamed with indentation preserved" };
+    },
+  },
+  {
+    name: "18. Insert blank line between sections",
+    fileName: "sections.txt",
+    fileContent: ["[section-a]", "value-a=1", "[section-b]", "value-b=2"].join("\n"),
+    prompt: [
+      "Read sections.txt with read_file.",
+      "Insert a blank empty line between 'value-a=1' (line 2) and '[section-b]' (line 3).",
+      "Use edit_file with op='append', pos=<line2 anchor>, lines=[''].",
+      "lines=[''] inserts one empty line.",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.split("\n");
+      const valAIdx = lines.findIndex((l) => l.includes("value-a=1"));
+      const secBIdx = lines.findIndex((l) => l.includes("[section-b]"));
+      if (valAIdx === -1) return { passed: false, reason: "'value-a=1' missing" };
+      if (secBIdx === -1) return { passed: false, reason: "'[section-b]' missing" };
+      if (secBIdx - valAIdx < 2) return { passed: false, reason: "no blank line between sections" };
+      const between = lines[valAIdx + 1];
+      if (between.trim() !== "") return { passed: false, reason: `line between is '${between}', not blank` };
+      return { passed: true, reason: "blank line inserted between sections" };
+    },
+  },
+  {
+    name: "19. Delete blank line",
+    fileName: "noblank.txt",
+    fileContent: ["first", "", "second", "third"].join("\n"),
+    prompt: [
+      "Read noblank.txt with read_file.",
+      "Delete the empty blank line (line 2). Use edit_file with op='replace', pos=<line2 anchor>, lines=[].",
+    ].join(" "),
+    validate: (c) => {
+      const lines = c.trim().split("\n");
+      if (lines.length !== 3) return { passed: false, reason: `expected 3 lines, got ${lines.length}` };
+      if (lines[0].trim() !== "first") return { passed: false, reason: "'first' not on line 1" };
+      if (lines[1].trim() !== "second") return { passed: false, reason: "'second' not on line 2" };
+      return { passed: true, reason: "blank line deleted" };
+    },
+  },
+  {
+    name: "20. Tab → spaces conversion",
+    fileName: "tabs.txt",
+    fileContent: ["start", "\tindented-with-tab", "end"].join("\n"),
+    prompt: [
+      "Read tabs.txt with read_file.",
+      "Replace the tab-indented line 2 using edit_file with edits: [{ op: 'replace', pos: '<line2 anchor>', lines: ['    indented-with-spaces'] }].",
+      "Expected final line 2 to be 4 spaces followed by indented-with-spaces.",
+    ].join(" "),
+    validate: (c) => {
+      if (c.includes("\t")) return { passed: false, reason: "tab still present" };
+      if (!c.includes("    indented-with-spaces"))
+        return { passed: false, reason: "'    indented-with-spaces' not found" };
+      if (!c.includes("start")) return { passed: false, reason: "'start' was modified" };
+      return { passed: true, reason: "tab converted to 4 spaces" };
+    },
+  },
+  {
+    name: "21. Deeply nested indent replacement",
+    fileName: "nested.ts",
+    fileContent: [
+      "if (a) {",
+      "  if (b) {",
+      "    if (c) {",
+      "      old_call();",
+      "    }",
+      "  }",
+      "}",
+    ].join("\n"),
+    prompt: [
+      "Read nested.ts with read_file.",
+      "Replace line 4 '      old_call();' with '      new_call();'.",
+      "Preserve the exact 6-space indentation. Only change the function name.",
+      "Use edit_file with op='replace', pos=<line4 anchor>, lines=['      new_call();'].",
+    ].join(" "),
+    validate: (c) => {
+      if (c.includes("old_call")) return { passed: false, reason: "'old_call' still present" };
+      const lines = c.split("\n");
+      const callLine = lines.find((l) => l.includes("new_call"));
+      if (!callLine) return { passed: false, reason: "'new_call' not found" };
+      const leadingSpaces = callLine.match(/^ */)?.[0].length ?? 0;
+      if (leadingSpaces !== 6) return { passed: false, reason: `expected 6-space indent, got ${leadingSpaces}` };
+      return { passed: true, reason: "deeply nested line replaced with indent preserved" };
+    },
+  },
+];
+
+// ── JSONL event types ─────────────────────────────────────────
+interface ToolCallEvent {
+  tool_call_id: string;
+  tool_input: Record<string, unknown>;
+  tool_name: string;
+  type: "tool_call";
+}
+
+interface ToolResultEvent {
+  error?: string;
+  output: string;
+  tool_call_id: string;
+  type: "tool_result";
+}
+
+interface AnyEvent {
+  type: string;
+  [key: string]: unknown;
+}
+
+// ── Run single test case ─────────────────────────────────────
+async function runTestCase(
+  tc: TestCase,
+  testDir: string
+): Promise<{
+  passed: boolean;
+  editCalls: number;
+  editSuccesses: number;
+  duration: number;
+}> {
+  const testFile = join(testDir, tc.fileName);
+  writeFileSync(testFile, tc.fileContent, "utf-8");
+
+  const headlessScript = resolve(import.meta.dir, "headless.ts");
+  const headlessArgs = [
+    "run",
+    headlessScript,
+    "-p",
+    tc.prompt,
+    "--no-translate",
+    ...extraArgs,
+  ];
+
+  const startTime = Date.now();
+
+  const output = await new Promise<string>((res, reject) => {
+    const proc = spawn("bun", headlessArgs, {
+      cwd: testDir,
+      env: { ...process.env, BUN_INSTALL: process.env.BUN_INSTALL },
+      stdio: ["ignore", "pipe", "pipe"],
+    });
+
+    let stdout = "";
+    let stderr = "";
+
+    proc.stdout.on("data", (chunk: Buffer) => {
+      stdout += chunk.toString();
+    });
+    proc.stderr.on("data", (chunk: Buffer) => {
+      stderr += chunk.toString();
+    });
+
+    const timeout = setTimeout(
+      () => {
+        proc.kill("SIGTERM");
+        reject(new Error("Timed out after 4 minutes"));
+      },
+      4 * 60 * 1000
+    );
+
+    proc.on("close", (code) => {
+      clearTimeout(timeout);
+      if (code !== 0) {
+        reject(new Error(`Exit code ${code}\n${stderr.slice(-500)}`));
+      } else {
+        res(stdout);
+      }
+    });
+    proc.on("error", (err) => {
+      clearTimeout(timeout);
+      reject(err);
+    });
+  });
+
+  const duration = Date.now() - startTime;
+
+  // Parse events
+  const events: AnyEvent[] = [];
+  for (const line of output.split("\n").filter((l) => l.trim())) {
+    try {
+      events.push(JSON.parse(line) as AnyEvent);
+    } catch {
+      // skip non-JSON
+    }
+  }
+
+  const toolCalls = events.filter(
+    (e) => e.type === "tool_call"
+  ) as unknown as ToolCallEvent[];
+  const toolResults = events.filter(
+    (e) => e.type === "tool_result"
+  ) as unknown as ToolResultEvent[];
+
+  const editCalls = toolCalls.filter((e) => e.tool_name === "edit_file");
+  const editCallIds = new Set(editCalls.map((e) => e.tool_call_id));
+  const editResults = toolResults.filter((e) =>
+    editCallIds.has(e.tool_call_id)
+  );
+  const editSuccesses = editResults.filter((e) => !e.error);
+
+  // Show blocked calls
+  const editErrors = editResults.filter((e) => e.error);
+  for (const err of editErrors) {
+    const matchingCall = editCalls.find(
+      (c) => c.tool_call_id === err.tool_call_id
+    );
+    info(`  blocked: ${err.error?.slice(0, 120)}`);
+    if (matchingCall) {
+      info(`  input: ${JSON.stringify(matchingCall.tool_input).slice(0, 200)}`);
+    }
+  }
+
+  // Validate file content
+  let finalContent: string;
+  try {
+    finalContent = readFileSync(testFile, "utf-8");
+  } catch {
+    return {
+      passed: false,
+      editCalls: editCalls.length,
+      editSuccesses: editSuccesses.length,
+      duration,
+    };
+  }
+
+  const validation = tc.validate(finalContent);
+
+  return {
+    passed: validation.passed,
+    editCalls: editCalls.length,
+    editSuccesses: editSuccesses.length,
+    duration,
+  };
+}
+
+// ── Main ──────────────────────────────────────────────────────
+const main = async () => {
+  console.log(`\n${BOLD}Headless Edit Operations Test — ${TEST_CASES.length} Types${RESET}\n`);
+
+  const testDir = join(tmpdir(), `edit-ops-${Date.now()}`);
+  mkdirSync(testDir, { recursive: true });
+  info(`Test dir: ${testDir}`);
+  console.log();
+
+  let totalPassed = 0;
+  const results: { name: string; passed: boolean; detail: string }[] = [];
+
+  for (const tc of TEST_CASES) {
+    console.log(`${CYAN}${BOLD}${tc.name}${RESET}`);
+    info(`File: ${tc.fileName}`);
+    info(`Prompt: "${tc.prompt.slice(0, 80)}..."`);
+
+    try {
+      const result = await runTestCase(tc, testDir);
+      const status = result.passed
+        ? `${GREEN}PASS${RESET}`
+        : `${RED}FAIL${RESET}`;
+      const detail = `edit_file: ${result.editSuccesses}/${result.editCalls} succeeded, ${(result.duration / 1000).toFixed(1)}s`;
+
+      console.log(`  ${status} — ${detail}`);
+
+      if (result.passed) {
+        totalPassed++;
+        // Validate the file to show reason
+        const content = readFileSync(join(testDir, tc.fileName), "utf-8");
+        const v = tc.validate(content);
+        pass(v.reason);
+      } else {
+        const content = readFileSync(join(testDir, tc.fileName), "utf-8");
+        const v = tc.validate(content);
+        fail(v.reason);
+        info(
+          `Final content:\n${content
+            .split("\n")
+            .map((l, i) => `    ${i + 1}: ${l}`)
+            .join("\n")}`
+        );
+      }
+
+      results.push({ name: tc.name, passed: result.passed, detail });
+    } catch (error) {
+      const msg = error instanceof Error ? error.message : String(error);
+      console.log(`  ${RED}ERROR${RESET} — ${msg.slice(0, 200)}`);
+      fail(msg.slice(0, 200));
+      results.push({ name: tc.name, passed: false, detail: msg.slice(0, 100) });
+    }
+
+    // Reset file for next test (in case of side effects)
+    try {
+      rmSync(join(testDir, tc.fileName), { force: true });
+    } catch {}
+
+    console.log();
+  }
+
+  // Summary
+  console.log(`${BOLD}━━━ Summary ━━━${RESET}`);
+  for (const r of results) {
+    const icon = r.passed ? `${GREEN}✓${RESET}` : `${RED}✗${RESET}`;
+    console.log(`  ${icon} ${r.name} — ${r.detail}`);
+  }
+  console.log();
+  console.log(
+    `${BOLD}Result: ${totalPassed}/${TEST_CASES.length} passed (${Math.round((totalPassed / TEST_CASES.length) * 100)}%)${RESET}`
+  );
+
+  // Cleanup
+  try {
+    rmSync(testDir, { recursive: true, force: true });
+  } catch {}
+
+  if (totalPassed === TEST_CASES.length) {
+    console.log(
+      `\n${BOLD}${GREEN}🎉 ALL TESTS PASSED — 100% success rate!${RESET}\n`
+    );
+    process.exit(0);
+  } else {
+    console.log(`\n${BOLD}${RED}Some tests failed.${RESET}\n`);
+    process.exit(1);
+  }
+};
+
+main();

benchmarks/test-multi-model.ts

@@ -0,0 +1,280 @@
+#!/usr/bin/env bun
+/**
+ * Multi-model edit_file test runner
+ *
+ * Runs test-headless-edit-ops.ts against every available model
+ * and produces a summary table.
+ *
+ * Usage:
+ *   bun run scripts/test-multi-model-edit.ts [--timeout <seconds>]
+ */
+
+import { spawn } from "node:child_process";
+import { resolve } from "node:path";
+
+// ── Models ────────────────────────────────────────────────────
+const MODELS = [
+  { id: "MiniMaxAI/MiniMax-M2.5", short: "M2.5" },
+  // { id: "MiniMaxAI/MiniMax-M2.1", short: "M2.1" },  // masked: slow + timeout-prone
+  // { id: "zai-org/GLM-5", short: "GLM-5" },            // masked: API 503
+  { id: "zai-org/GLM-4.7", short: "GLM-4.7" },
+];
+
+// ── CLI args ──────────────────────────────────────────────────
+let perModelTimeoutSec = 900; // 15 min default per model (5 tests)
+const rawArgs = process.argv.slice(2);
+for (let i = 0; i < rawArgs.length; i++) {
+  if (rawArgs[i] === "--timeout" && i + 1 < rawArgs.length) {
+    const parsed = Number.parseInt(rawArgs[i + 1], 10);
+    if (Number.isNaN(parsed) || parsed <= 0) {
+      console.error(`Invalid --timeout value: ${rawArgs[i + 1]}`);
+      process.exit(1);
+    }
+    perModelTimeoutSec = parsed;
+    i++;
+}
+
+// ── Colors ────────────────────────────────────────────────────
+const BOLD = "\x1b[1m";
+const GREEN = "\x1b[32m";
+const RED = "\x1b[31m";
+const YELLOW = "\x1b[33m";
+const DIM = "\x1b[2m";
+const CYAN = "\x1b[36m";
+const RESET = "\x1b[0m";
+
+// ── Types ─────────────────────────────────────────────────────
+interface TestResult {
+  detail: string;
+  name: string;
+  passed: boolean;
+}
+
+interface ModelResult {
+  durationMs: number;
+  error?: string;
+  modelId: string;
+  modelShort: string;
+  tests: TestResult[];
+  totalPassed: number;
+  totalTests: number;
+}
+
+// ── Parse test-headless-edit-ops stdout ───────────────────────
+function parseOpsOutput(stdout: string): TestResult[] {
+  const results: TestResult[] = [];
+
+  // Match lines like: "  PASS — edit_file: 1/1 succeeded, 32.5s"
+  // or "  FAIL — edit_file: 0/3 succeeded, 15.2s"
+  // or "  ERROR — Timed out after 10 minutes"
+  // Following a line like: "1. Replace single line"
+  const lines = stdout.split("\n");
+
+  let currentTestName = "";
+  for (const line of lines) {
+    // Detect test name: starts with ANSI-colored bold cyan + "N. Name"
+    // Strip ANSI codes for matching
+    const stripped = line.replace(/\x1b\[[0-9;]*m/g, "");
+
+    // Test name pattern: "N. <name>"
+    const testNameMatch = stripped.match(/^\s*(\d+\.\s+.+)$/);
+    if (
+      testNameMatch &&
+      !stripped.includes("—") &&
+      !stripped.includes("✓") &&
+      !stripped.includes("✗")
+    ) {
+      currentTestName = testNameMatch[1].trim();
+      continue;
+    }
+
+    // Result line: PASS/FAIL/ERROR
+    if (currentTestName && stripped.includes("PASS")) {
+      const detail = stripped.replace(/^\s*PASS\s*—?\s*/, "").trim();
+      results.push({
+        name: currentTestName,
+        passed: true,
+        detail: detail || "passed",
+      });
+      currentTestName = "";
+    } else if (currentTestName && stripped.includes("FAIL")) {
+      const detail = stripped.replace(/^\s*FAIL\s*—?\s*/, "").trim();
+      results.push({
+        name: currentTestName,
+        passed: false,
+        detail: detail || "failed",
+      });
+      currentTestName = "";
+    } else if (currentTestName && stripped.includes("ERROR")) {
+      const detail = stripped.replace(/^\s*ERROR\s*—?\s*/, "").trim();
+      results.push({
+        name: currentTestName,
+        passed: false,
+        detail: detail || "error",
+      });
+      currentTestName = "";
+    }
+  }
+
+  return results;
+}
+
+// ── Run one model ────────────────────────────────────────────
+async function runModel(model: {
+  id: string;
+  short: string;
+}): Promise<ModelResult> {
+  const opsScript = resolve(import.meta.dir, "test-edit-ops.ts");
+  const startTime = Date.now();
+
+  return new Promise<ModelResult>((resolvePromise) => {
+    const proc = spawn(
+      "bun",
+      ["run", opsScript, "-m", model.id, "--no-translate"],
+      {
+        cwd: resolve(import.meta.dir),
+        env: { ...process.env, BUN_INSTALL: process.env.BUN_INSTALL },
+        stdio: ["ignore", "pipe", "pipe"],
+      }
+    );
+
+    let stdout = "";
+    let stderr = "";
+
+    proc.stdout.on("data", (chunk: Buffer) => {
+      stdout += chunk.toString();
+    });
+    proc.stderr.on("data", (chunk: Buffer) => {
+      stderr += chunk.toString();
+    });
+
+    const timeout = setTimeout(() => {
+      proc.kill("SIGTERM");
+      resolvePromise({
+        modelId: model.id,
+        modelShort: model.short,
+        tests: [],
+        totalPassed: 0,
+        totalTests: 0,
+        durationMs: Date.now() - startTime,
+        error: `Timed out after ${perModelTimeoutSec}s`,
+      });
+    }, perModelTimeoutSec * 1000);
+
+    proc.on("close", () => {
+      clearTimeout(timeout);
+      const tests = parseOpsOutput(stdout);
+      const totalPassed = tests.filter((t) => t.passed).length;
+
+      resolvePromise({
+        modelId: model.id,
+        modelShort: model.short,
+        tests,
+        totalPassed,
+        totalTests: Math.max(tests.length, 5),
+        durationMs: Date.now() - startTime,
+      });
+    });
+
+    proc.on("error", (err) => {
+      clearTimeout(timeout);
+      resolvePromise({
+        modelId: model.id,
+        modelShort: model.short,
+        tests: [],
+        totalPassed: 0,
+        totalTests: 0,
+        durationMs: Date.now() - startTime,
+        error: err.message,
+      });
+    });
+  });
+}
+
+// ── Main ──────────────────────────────────────────────────────
+const main = async () => {
+  console.log(`\n${BOLD}═══ Multi-Model edit_file Test Runner ═══${RESET}\n`);
+  console.log(`${DIM}Models: ${MODELS.map((m) => m.short).join(", ")}${RESET}`);
+  console.log(`${DIM}Timeout: ${perModelTimeoutSec}s per model${RESET}`);
+  console.log();
+
+  const allResults: ModelResult[] = [];
+
+  for (const model of MODELS) {
+    console.log(`${CYAN}${BOLD}▶ Testing ${model.short} (${model.id})${RESET}`);
+    const result = await runModel(model);
+    allResults.push(result);
+
+    const timeStr = `${(result.durationMs / 1000).toFixed(1)}s`;
+    if (result.error) {
+      console.log(`  ${RED}ERROR${RESET}: ${result.error} (${timeStr})`);
+    } else {
+      const color =
+        result.totalPassed === result.totalTests
+          ? GREEN
+          : result.totalPassed > 0
+            ? YELLOW
+            : RED;
+      console.log(
+        `  ${color}${result.totalPassed}/${result.totalTests} passed${RESET} (${timeStr})`
+      );
+      for (const t of result.tests) {
+        const icon = t.passed ? `${GREEN}✓${RESET}` : `${RED}✗${RESET}`;
+        console.log(`    ${icon} ${t.name}`);
+      }
+    }
+    console.log();
+  }
+
+  // ── Summary Table ──────────────────────────────────────────
+  console.log(`${BOLD}═══ Summary ═══${RESET}\n`);
+
+  // Per-model results
+  for (const r of allResults) {
+    const timeStr = `${(r.durationMs / 1000).toFixed(0)}s`;
+    const color = r.error ? RED : r.totalPassed === r.totalTests ? GREEN : r.totalPassed > 0 ? YELLOW : RED;
+    const label = r.error ? `ERROR: ${r.error}` : `${r.totalPassed}/${r.totalTests}`;
+    console.log(`  ${r.modelShort.padEnd(8)} ${color}${label}${RESET} (${timeStr})`);
+    for (const t of r.tests) {
+      const icon = t.passed ? `${GREEN}✓${RESET}` : `${RED}✗${RESET}`;
+      console.log(`    ${icon} ${t.name}`);
+    }
+  }
+
+  console.log();
+
+  // Overall
+  const totalModels = allResults.length;
+  const erroredModels = allResults.filter((r) => r.error).length;
+  const perfectModels = allResults.filter(
+    (r) => !r.error && r.totalPassed === r.totalTests && r.totalTests > 0
+  ).length;
+  console.log(
+    `${BOLD}Models with 100%: ${perfectModels}/${totalModels}${RESET}`
+  );
+
+  const overallPassed = allResults.reduce((sum, r) => sum + r.totalPassed, 0);
+  const overallTotal = allResults.reduce((sum, r) => sum + r.totalTests, 0);
+  console.log(
+    `${BOLD}Overall: ${overallPassed}/${overallTotal} (${Math.round((overallPassed / overallTotal) * 100)}%)${RESET}`
+  );
+
+  console.log();
+
+  if (erroredModels > 0) {
+    console.log(
+      `${BOLD}${RED}${erroredModels} model(s) errored. See details above.${RESET}\n`
+    );
+    process.exit(1);
+  } else if (perfectModels === totalModels) {
+    console.log(`${BOLD}${GREEN}🎉 ALL MODELS PASSED ALL TESTS!${RESET}\n`);
+    process.exit(0);
+  } else {
+    console.log(
+      `${BOLD}${YELLOW}Some models have failures. See details above.${RESET}\n`
+    );
+    process.exit(1);
+  }
+};
+
+main();

bin/oh-my-opencode.js

@@ -3,8 +3,9 @@
 // Wrapper script that detects platform and spawns the correct binary
 
 import { spawnSync } from "node:child_process";
+import { readFileSync } from "node:fs";
 import { createRequire } from "node:module";
-import { getPlatformPackage, getBinaryPath } from "./platform.js";
+import { getPlatformPackageCandidates, getBinaryPath } from "./platform.js";
 
 const require = createRequire(import.meta.url);
 
@@ -26,55 +27,116 @@ function getLibcFamily() {
   }
 }
 
+function supportsAvx2() {
+  if (process.arch !== "x64") {
+    return null;
+  }
+
+  if (process.env.OH_MY_OPENCODE_FORCE_BASELINE === "1") {
+    return false;
+  }
+
+  if (process.platform === "linux") {
+    try {
+      const cpuInfo = readFileSync("/proc/cpuinfo", "utf8").toLowerCase();
+      return cpuInfo.includes("avx2");
+    } catch {
+      return null;
+    }
+  }
+
+  if (process.platform === "darwin") {
+    const probe = spawnSync("sysctl", ["-n", "machdep.cpu.leaf7_features"], {
+      encoding: "utf8",
+    });
+
+    if (probe.error || probe.status !== 0) {
+      return null;
+    }
+
+    return probe.stdout.toUpperCase().includes("AVX2");
+  }
+
+  return null;
+}
+
+function getSignalExitCode(signal) {
+  const signalCodeByName = {
+    SIGINT: 2,
+    SIGILL: 4,
+    SIGKILL: 9,
+    SIGTERM: 15,
+  };
+
+  return 128 + (signalCodeByName[signal] ?? 1);
+}
+
 function main() {
   const { platform, arch } = process;
   const libcFamily = getLibcFamily();
+  const avx2Supported = supportsAvx2();
   
-  // Get platform package name
-  let pkg;
+  let packageCandidates;
   try {
-    pkg = getPlatformPackage({ platform, arch, libcFamily });
+    packageCandidates = getPlatformPackageCandidates({
+      platform,
+      arch,
+      libcFamily,
+      preferBaseline: avx2Supported === false,
+    });
   } catch (error) {
     console.error(`\noh-my-opencode: ${error.message}\n`);
     process.exit(1);
   }
-  
-  // Resolve binary path
-  const binRelPath = getBinaryPath(pkg, platform);
-  
-  let binPath;
-  try {
-    binPath = require.resolve(binRelPath);
-  } catch {
+
+  const resolvedBinaries = packageCandidates
+    .map((pkg) => {
+      try {
+        return { pkg, binPath: require.resolve(getBinaryPath(pkg, platform)) };
+      } catch {
+        return null;
+      }
+    })
+    .filter((entry) => entry !== null);
+
+  if (resolvedBinaries.length === 0) {
     console.error(`\noh-my-opencode: Platform binary not installed.`);
     console.error(`\nYour platform: ${platform}-${arch}${libcFamily === "musl" ? "-musl" : ""}`);
-    console.error(`Expected package: ${pkg}`);
+    console.error(`Expected packages (in order): ${packageCandidates.join(", ")}`);
     console.error(`\nTo fix, run:`);
-    console.error(`  npm install ${pkg}\n`);
+    console.error(`  npm install ${packageCandidates[0]}\n`);
     process.exit(1);
   }
-  
-  // Spawn the binary
-  const result = spawnSync(binPath, process.argv.slice(2), {
-    stdio: "inherit",
-  });
-  
-  // Handle spawn errors
-  if (result.error) {
-    console.error(`\noh-my-opencode: Failed to execute binary.`);
-    console.error(`Error: ${result.error.message}\n`);
-    process.exit(2);
-  }
-  
-  // Handle signals
-  if (result.signal) {
-    const signalNum = result.signal === "SIGTERM" ? 15 : 
-                      result.signal === "SIGKILL" ? 9 :
-                      result.signal === "SIGINT" ? 2 : 1;
-    process.exit(128 + signalNum);
+
+  for (let index = 0; index < resolvedBinaries.length; index += 1) {
+    const currentBinary = resolvedBinaries[index];
+    const hasFallback = index < resolvedBinaries.length - 1;
+    const result = spawnSync(currentBinary.binPath, process.argv.slice(2), {
+      stdio: "inherit",
+    });
+
+    if (result.error) {
+      if (hasFallback) {
+        continue;
+      }
+
+      console.error(`\noh-my-opencode: Failed to execute binary.`);
+      console.error(`Error: ${result.error.message}\n`);
+      process.exit(2);
+    }
+
+    if (result.signal === "SIGILL" && hasFallback) {
+      continue;
+    }
+
+    if (result.signal) {
+      process.exit(getSignalExitCode(result.signal));
+    }
+
+    process.exit(result.status ?? 1);
   }
 
-  process.exit(result.status ?? 1);
+  process.exit(1);
 }
 
 main();

bin/platform.d.ts

@@ -0,0 +1,14 @@
+export declare function getPlatformPackage(options: {
+  platform: string;
+  arch: string;
+  libcFamily?: string | null;
+}): string;
+
+export declare function getPlatformPackageCandidates(options: {
+  platform: string;
+  arch: string;
+  libcFamily?: string | null;
+  preferBaseline?: boolean;
+}): string[];
+
+export declare function getBinaryPath(pkg: string, platform: string): string;

bin/platform.js

@@ -26,6 +26,50 @@ export function getPlatformPackage({ platform, arch, libcFamily }) {
   return `oh-my-opencode-${os}-${arch}${suffix}`;
 }
 
+/** @param {{ platform: string, arch: string, libcFamily?: string | null, preferBaseline?: boolean }} options */
+export function getPlatformPackageCandidates({ platform, arch, libcFamily, preferBaseline = false }) {
+  const primaryPackage = getPlatformPackage({ platform, arch, libcFamily });
+  const baselinePackage = getBaselinePlatformPackage({ platform, arch, libcFamily });
+
+  if (!baselinePackage) {
+    return [primaryPackage];
+  }
+
+  return preferBaseline ? [baselinePackage, primaryPackage] : [primaryPackage, baselinePackage];
+}
+
+/** @param {{ platform: string, arch: string, libcFamily?: string | null }} options */
+function getBaselinePlatformPackage({ platform, arch, libcFamily }) {
+  if (arch !== "x64") {
+    return null;
+  }
+
+  if (platform === "darwin") {
+    return "oh-my-opencode-darwin-x64-baseline";
+  }
+
+  if (platform === "win32") {
+    return "oh-my-opencode-windows-x64-baseline";
+  }
+
+  if (platform === "linux") {
+    if (libcFamily === null || libcFamily === undefined) {
+      throw new Error(
+        "Could not detect libc on Linux. " +
+        "Please ensure detect-libc is installed or report this issue."
+      );
+    }
+
+    if (libcFamily === "musl") {
+      return "oh-my-opencode-linux-x64-musl-baseline";
+    }
+
+    return "oh-my-opencode-linux-x64-baseline";
+  }
+
+  return null;
+}
+
 /**
  * Get the path to the binary within a platform package
  * @param {string} pkg Package name

bin/platform.test.ts

@@ -1,6 +1,6 @@
 // bin/platform.test.ts
 import { describe, expect, test } from "bun:test";
-import { getPlatformPackage, getBinaryPath } from "./platform.js";
+import { getBinaryPath, getPlatformPackage, getPlatformPackageCandidates } from "./platform.js";
 
 describe("getPlatformPackage", () => {
   // #region Darwin platforms
@@ -146,3 +146,58 @@ describe("getBinaryPath", () => {
     expect(result).toBe("oh-my-opencode-linux-x64/bin/oh-my-opencode");
   });
 });
+
+describe("getPlatformPackageCandidates", () => {
+  test("returns x64 and baseline candidates for Linux glibc", () => {
+    // #given Linux x64 with glibc
+    const input = { platform: "linux", arch: "x64", libcFamily: "glibc" };
+
+    // #when getting package candidates
+    const result = getPlatformPackageCandidates(input);
+
+    // #then returns modern first then baseline fallback
+    expect(result).toEqual([
+      "oh-my-opencode-linux-x64",
+      "oh-my-opencode-linux-x64-baseline",
+    ]);
+  });
+
+  test("returns x64 musl and baseline candidates for Linux musl", () => {
+    // #given Linux x64 with musl
+    const input = { platform: "linux", arch: "x64", libcFamily: "musl" };
+
+    // #when getting package candidates
+    const result = getPlatformPackageCandidates(input);
+
+    // #then returns musl modern first then musl baseline fallback
+    expect(result).toEqual([
+      "oh-my-opencode-linux-x64-musl",
+      "oh-my-opencode-linux-x64-musl-baseline",
+    ]);
+  });
+
+  test("returns baseline first when preferBaseline is true", () => {
+    // #given Windows x64 and baseline preference
+    const input = { platform: "win32", arch: "x64", preferBaseline: true };
+
+    // #when getting package candidates
+    const result = getPlatformPackageCandidates(input);
+
+    // #then baseline package is preferred first
+    expect(result).toEqual([
+      "oh-my-opencode-windows-x64-baseline",
+      "oh-my-opencode-windows-x64",
+    ]);
+  });
+
+  test("returns only one candidate for ARM64", () => {
+    // #given non-x64 platform
+    const input = { platform: "linux", arch: "arm64", libcFamily: "glibc" };
+
+    // #when getting package candidates
+    const result = getPlatformPackageCandidates(input);
+
+    // #then baseline fallback is not included
+    expect(result).toEqual(["oh-my-opencode-linux-arm64"]);
+  });
+});

package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "The Best AI Agent Harness - Batteries-Included OpenCode Plugin with Multi-Model Orchestration, Parallel Background Agents, and Crafted LSP/AST Tools",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -75,13 +75,17 @@
     "typescript": "^5.7.3"
   },
   "optionalDependencies": {
-    "oh-my-opencode-darwin-arm64": "3.8.5",
-    "oh-my-opencode-darwin-x64": "3.8.5",
-    "oh-my-opencode-linux-arm64": "3.8.5",
-    "oh-my-opencode-linux-arm64-musl": "3.8.5",
-    "oh-my-opencode-linux-x64": "3.8.5",
-    "oh-my-opencode-linux-x64-musl": "3.8.5",
-    "oh-my-opencode-windows-x64": "3.8.5"
+    "oh-my-opencode-darwin-arm64": "3.9.0",
+    "oh-my-opencode-darwin-x64": "3.9.0",
+    "oh-my-opencode-darwin-x64-baseline": "3.9.0",
+    "oh-my-opencode-linux-arm64": "3.9.0",
+    "oh-my-opencode-linux-arm64-musl": "3.9.0",
+    "oh-my-opencode-linux-x64": "3.9.0",
+    "oh-my-opencode-linux-x64-baseline": "3.9.0",
+    "oh-my-opencode-linux-x64-musl": "3.9.0",
+    "oh-my-opencode-linux-x64-musl-baseline": "3.9.0",
+    "oh-my-opencode-windows-x64": "3.9.0",
+    "oh-my-opencode-windows-x64-baseline": "3.9.0"
   },
   "trustedDependencies": [
     "@ast-grep/cli",

packages/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-darwin-arm64",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (darwin-arm64)",
   "license": "MIT",
   "repository": {

packages/darwin-x64-baseline/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-darwin-x64-baseline",
-  "version": "3.1.1",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (darwin-x64-baseline, no AVX2)",
   "license": "MIT",
   "repository": {

packages/darwin-x64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-darwin-x64",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (darwin-x64)",
   "license": "MIT",
   "repository": {

packages/linux-arm64-musl/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-arm64-musl",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (linux-arm64-musl)",
   "license": "MIT",
   "repository": {

packages/linux-arm64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-arm64",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (linux-arm64)",
   "license": "MIT",
   "repository": {

packages/linux-x64-baseline/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-x64-baseline",
-  "version": "3.1.1",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (linux-x64-baseline, no AVX2)",
   "license": "MIT",
   "repository": {

packages/linux-x64-musl-baseline/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-x64-musl-baseline",
-  "version": "3.1.1",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (linux-x64-musl-baseline, no AVX2)",
   "license": "MIT",
   "repository": {

packages/linux-x64-musl/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-x64-musl",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (linux-x64-musl)",
   "license": "MIT",
   "repository": {

packages/linux-x64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-x64",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (linux-x64)",
   "license": "MIT",
   "repository": {

packages/windows-x64-baseline/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-windows-x64-baseline",
-  "version": "3.1.1",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (windows-x64-baseline, no AVX2)",
   "license": "MIT",
   "repository": {

packages/windows-x64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-windows-x64",
-  "version": "3.8.5",
+  "version": "3.9.0",
   "description": "Platform-specific binary for oh-my-opencode (windows-x64)",
   "license": "MIT",
   "repository": {

postinstall.mjs

@@ -2,7 +2,7 @@
 // Runs after npm install to verify platform binary is available
 
 import { createRequire } from "node:module";
-import { getPlatformPackage, getBinaryPath } from "./bin/platform.js";
+import { getPlatformPackageCandidates, getBinaryPath } from "./bin/platform.js";
 
 const require = createRequire(import.meta.url);
 
@@ -27,12 +27,28 @@ function main() {
   const libcFamily = getLibcFamily();
   
   try {
-    const pkg = getPlatformPackage({ platform, arch, libcFamily });
-    const binPath = getBinaryPath(pkg, platform);
-    
-    // Try to resolve the binary
-    require.resolve(binPath);
-    console.log(`✓ oh-my-opencode binary installed for ${platform}-${arch}`);
+    const packageCandidates = getPlatformPackageCandidates({
+      platform,
+      arch,
+      libcFamily,
+    });
+
+    const resolvedPackage = packageCandidates.find((pkg) => {
+      try {
+        require.resolve(getBinaryPath(pkg, platform));
+        return true;
+      } catch {
+        return false;
+      }
+    });
+
+    if (!resolvedPackage) {
+      throw new Error(
+        `No platform binary package installed. Tried: ${packageCandidates.join(", ")}`
+      );
+    }
+
+    console.log(`✓ oh-my-opencode binary installed for ${platform}-${arch} (${resolvedPackage})`);
   } catch (error) {
     console.warn(`⚠ oh-my-opencode: ${error.message}`);
     console.warn(`  The CLI may not work on this platform.`);

src/agents/atlas/agent.ts

@@ -29,7 +29,7 @@ import {
   buildDecisionMatrix,
 } from "./prompt-section-builder"
 
-const MODE: AgentMode = "primary"
+const MODE: AgentMode = "all"
 
 export type AtlasPromptSource = "default" | "gpt" | "gemini"
 

src/agents/hephaestus.ts

@@ -19,7 +19,7 @@ import {
   categorizeTools,
 } from "./dynamic-agent-prompt-builder";
 
-const MODE: AgentMode = "primary";
+const MODE: AgentMode = "all";
 
 function buildTodoDisciplineSection(useTaskSystem: boolean): string {
   if (useTaskSystem) {
@@ -448,21 +448,6 @@ ${oracleSection}
 4. **Run build** if applicable — exit code 0 required
 5. **Tell user** what you verified and the results — keep it clear and helpful
 
-### Auto-Commit Policy (MANDATORY for implementation/fix work)
-
-1. **Auto-commit after implementation is complete** when the task includes feature/fix code changes
-2. **Commit ONLY after verification gates pass**:
-   - \`lsp_diagnostics\` clean on all modified files
-   - Related tests pass
-   - Typecheck/build pass when applicable
-3. **If any gate fails, DO NOT commit** — fix issues first, re-run verification, then commit
-4. **Use Conventional Commits format** with meaningful intent-focused messages:
-   - \`feat(scope): add ...\` for new functionality
-   - \`fix(scope): resolve ...\` for bug fixes
-   - \`refactor(scope): simplify ...\` for internal restructuring
-5. **Do not make placeholder commits** (\`wip\`, \`temp\`, \`update\`) or commit unverified code
-6. **If user explicitly says not to commit**, skip commit and report that changes are left uncommitted
-
 - **File edit** — \`lsp_diagnostics\` clean
 - **Build** — Exit code 0
 - **Tests** — Pass (or pre-existing failures noted)

src/agents/sisyphus.ts

@@ -8,7 +8,7 @@ import {
   buildGeminiIntentGateEnforcement,
 } from "./sisyphus-gemini-overlays";
 
-const MODE: AgentMode = "primary";
+const MODE: AgentMode = "all";
 export const SISYPHUS_PROMPT_METADATA: AgentPromptMetadata = {
   category: "utility",
   cost: "EXPENSIVE",

src/agents/types.test.ts

@@ -47,8 +47,8 @@ describe("isGptModel", () => {
   });
 
   test("gemini models are not gpt", () => {
-    expect(isGptModel("google/gemini-3-pro")).toBe(false);
-    expect(isGptModel("litellm/gemini-3-pro")).toBe(false);
+    expect(isGptModel("google/gemini-3.1-pro")).toBe(false);
+    expect(isGptModel("litellm/gemini-3.1-pro")).toBe(false);
   });
 
   test("opencode provider is not gpt", () => {
@@ -58,29 +58,29 @@ describe("isGptModel", () => {
 
 describe("isGeminiModel", () => {
   test("#given google provider models #then returns true", () => {
-    expect(isGeminiModel("google/gemini-3-pro")).toBe(true);
+    expect(isGeminiModel("google/gemini-3.1-pro")).toBe(true);
     expect(isGeminiModel("google/gemini-3-flash")).toBe(true);
     expect(isGeminiModel("google/gemini-2.5-pro")).toBe(true);
   });
 
   test("#given google-vertex provider models #then returns true", () => {
-    expect(isGeminiModel("google-vertex/gemini-3-pro")).toBe(true);
+    expect(isGeminiModel("google-vertex/gemini-3.1-pro")).toBe(true);
     expect(isGeminiModel("google-vertex/gemini-3-flash")).toBe(true);
   });
 
   test("#given github copilot gemini models #then returns true", () => {
-    expect(isGeminiModel("github-copilot/gemini-3-pro")).toBe(true);
+    expect(isGeminiModel("github-copilot/gemini-3.1-pro")).toBe(true);
     expect(isGeminiModel("github-copilot/gemini-3-flash")).toBe(true);
   });
 
   test("#given litellm proxied gemini models #then returns true", () => {
-    expect(isGeminiModel("litellm/gemini-3-pro")).toBe(true);
+    expect(isGeminiModel("litellm/gemini-3.1-pro")).toBe(true);
     expect(isGeminiModel("litellm/gemini-3-flash")).toBe(true);
     expect(isGeminiModel("litellm/gemini-2.5-pro")).toBe(true);
   });
 
   test("#given other proxied gemini models #then returns true", () => {
-    expect(isGeminiModel("custom-provider/gemini-3-pro")).toBe(true);
+    expect(isGeminiModel("custom-provider/gemini-3.1-pro")).toBe(true);
     expect(isGeminiModel("ollama/gemini-3-flash")).toBe(true);
   });
 

src/agents/utils.test.ts

@@ -988,7 +988,7 @@ describe("buildAgent with category and skills", () => {
     const agent = buildAgent(source["test-agent"], TEST_MODEL)
 
     // #then - category's built-in model is applied
-    expect(agent.model).toBe("google/gemini-3-pro")
+    expect(agent.model).toBe("google/gemini-3.1-pro")
   })
 
   test("agent with category and existing model keeps existing model", () => {

src/cli/__snapshots__/model-fallback.test.ts.snap

@@ -325,7 +325,7 @@ exports[`generateModelConfig single native provider uses Gemini models when only
   "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/master/assets/oh-my-opencode.schema.json",
   "agents": {
     "atlas": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
     },
     "explore": {
       "model": "opencode/gpt-5-nano",
@@ -334,34 +334,34 @@ exports[`generateModelConfig single native provider uses Gemini models when only
       "model": "opencode/glm-4.7-free",
     },
     "metis": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "momus": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "multimodal-looker": {
       "model": "google/gemini-3-flash-preview",
     },
     "oracle": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "prometheus": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
     },
   },
   "categories": {
     "artistry": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "quick": {
       "model": "google/gemini-3-flash-preview",
     },
     "ultrabrain": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "unspecified-high": {
@@ -371,7 +371,7 @@ exports[`generateModelConfig single native provider uses Gemini models when only
       "model": "google/gemini-3-flash-preview",
     },
     "visual-engineering": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -386,7 +386,7 @@ exports[`generateModelConfig single native provider uses Gemini models with isMa
   "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/master/assets/oh-my-opencode.schema.json",
   "agents": {
     "atlas": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
     },
     "explore": {
       "model": "opencode/gpt-5-nano",
@@ -395,44 +395,44 @@ exports[`generateModelConfig single native provider uses Gemini models with isMa
       "model": "opencode/glm-4.7-free",
     },
     "metis": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "momus": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "multimodal-looker": {
       "model": "google/gemini-3-flash-preview",
     },
     "oracle": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "prometheus": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
     },
   },
   "categories": {
     "artistry": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "quick": {
       "model": "google/gemini-3-flash-preview",
     },
     "ultrabrain": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "unspecified-high": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
     },
     "unspecified-low": {
       "model": "google/gemini-3-flash-preview",
     },
     "visual-engineering": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -485,7 +485,7 @@ exports[`generateModelConfig all native providers uses preferred models from fal
   },
   "categories": {
     "artistry": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "deep": {
@@ -506,7 +506,7 @@ exports[`generateModelConfig all native providers uses preferred models from fal
       "model": "anthropic/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -559,7 +559,7 @@ exports[`generateModelConfig all native providers uses preferred models with isM
   },
   "categories": {
     "artistry": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "deep": {
@@ -581,7 +581,7 @@ exports[`generateModelConfig all native providers uses preferred models with isM
       "model": "anthropic/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -634,7 +634,7 @@ exports[`generateModelConfig fallback providers uses OpenCode Zen models when on
   },
   "categories": {
     "artistry": {
-      "model": "opencode/gemini-3-pro",
+      "model": "opencode/gemini-3.1-pro",
       "variant": "high",
     },
     "deep": {
@@ -655,7 +655,7 @@ exports[`generateModelConfig fallback providers uses OpenCode Zen models when on
       "model": "opencode/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "opencode/gemini-3-pro",
+      "model": "opencode/gemini-3.1-pro",
       "variant": "high",
     },
     "writing": {
@@ -708,7 +708,7 @@ exports[`generateModelConfig fallback providers uses OpenCode Zen models with is
   },
   "categories": {
     "artistry": {
-      "model": "opencode/gemini-3-pro",
+      "model": "opencode/gemini-3.1-pro",
       "variant": "high",
     },
     "deep": {
@@ -730,7 +730,7 @@ exports[`generateModelConfig fallback providers uses OpenCode Zen models with is
       "model": "opencode/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "opencode/gemini-3-pro",
+      "model": "opencode/gemini-3.1-pro",
       "variant": "high",
     },
     "writing": {
@@ -779,14 +779,14 @@ exports[`generateModelConfig fallback providers uses GitHub Copilot models when
   },
   "categories": {
     "artistry": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "quick": {
       "model": "github-copilot/claude-haiku-4.5",
     },
     "ultrabrain": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "unspecified-high": {
@@ -796,7 +796,7 @@ exports[`generateModelConfig fallback providers uses GitHub Copilot models when
       "model": "github-copilot/claude-sonnet-4.5",
     },
     "visual-engineering": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -845,14 +845,14 @@ exports[`generateModelConfig fallback providers uses GitHub Copilot models with
   },
   "categories": {
     "artistry": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "quick": {
       "model": "github-copilot/claude-haiku-4.5",
     },
     "ultrabrain": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "unspecified-high": {
@@ -863,7 +863,7 @@ exports[`generateModelConfig fallback providers uses GitHub Copilot models with
       "model": "github-copilot/claude-sonnet-4.5",
     },
     "visual-engineering": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -1026,7 +1026,7 @@ exports[`generateModelConfig mixed provider scenarios uses Claude + OpenCode Zen
   },
   "categories": {
     "artistry": {
-      "model": "opencode/gemini-3-pro",
+      "model": "opencode/gemini-3.1-pro",
       "variant": "high",
     },
     "deep": {
@@ -1047,7 +1047,7 @@ exports[`generateModelConfig mixed provider scenarios uses Claude + OpenCode Zen
       "model": "anthropic/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "opencode/gemini-3-pro",
+      "model": "opencode/gemini-3.1-pro",
       "variant": "high",
     },
     "writing": {
@@ -1100,7 +1100,7 @@ exports[`generateModelConfig mixed provider scenarios uses OpenAI + Copilot comb
   },
   "categories": {
     "artistry": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "deep": {
@@ -1121,7 +1121,7 @@ exports[`generateModelConfig mixed provider scenarios uses OpenAI + Copilot comb
       "model": "github-copilot/claude-sonnet-4.5",
     },
     "visual-engineering": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -1217,7 +1217,7 @@ exports[`generateModelConfig mixed provider scenarios uses Gemini + Claude combi
       "model": "google/gemini-3-flash-preview",
     },
     "oracle": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "prometheus": {
@@ -1231,14 +1231,14 @@ exports[`generateModelConfig mixed provider scenarios uses Gemini + Claude combi
   },
   "categories": {
     "artistry": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "quick": {
       "model": "anthropic/claude-haiku-4-5",
     },
     "ultrabrain": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "unspecified-high": {
@@ -1248,7 +1248,7 @@ exports[`generateModelConfig mixed provider scenarios uses Gemini + Claude combi
       "model": "anthropic/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -1301,7 +1301,7 @@ exports[`generateModelConfig mixed provider scenarios uses all fallback provider
   },
   "categories": {
     "artistry": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "deep": {
@@ -1322,7 +1322,7 @@ exports[`generateModelConfig mixed provider scenarios uses all fallback provider
       "model": "github-copilot/claude-sonnet-4.5",
     },
     "visual-engineering": {
-      "model": "github-copilot/gemini-3-pro-preview",
+      "model": "github-copilot/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -1375,7 +1375,7 @@ exports[`generateModelConfig mixed provider scenarios uses all providers togethe
   },
   "categories": {
     "artistry": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "deep": {
@@ -1396,7 +1396,7 @@ exports[`generateModelConfig mixed provider scenarios uses all providers togethe
       "model": "anthropic/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {
@@ -1449,7 +1449,7 @@ exports[`generateModelConfig mixed provider scenarios uses all providers with is
   },
   "categories": {
     "artistry": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "deep": {
@@ -1471,7 +1471,7 @@ exports[`generateModelConfig mixed provider scenarios uses all providers with is
       "model": "anthropic/claude-sonnet-4-5",
     },
     "visual-engineering": {
-      "model": "google/gemini-3-pro-preview",
+      "model": "google/gemini-3.1-pro-preview",
       "variant": "high",
     },
     "writing": {

src/cli/config-manager.test.ts

@@ -178,7 +178,7 @@ describe("config-manager ANTIGRAVITY_PROVIDER_CONFIG", () => {
     expect(models).toBeTruthy()
 
     const required = [
-      "antigravity-gemini-3-pro",
+      "antigravity-gemini-3.1-pro",
       "antigravity-gemini-3-flash",
       "antigravity-claude-sonnet-4-6",
       "antigravity-claude-sonnet-4-6-thinking",
@@ -206,7 +206,7 @@ describe("config-manager ANTIGRAVITY_PROVIDER_CONFIG", () => {
     const models = (ANTIGRAVITY_PROVIDER_CONFIG as any).google.models as Record<string, any>
 
     // #when checking Gemini Pro variants
-    const pro = models["antigravity-gemini-3-pro"]
+    const pro = models["antigravity-gemini-3.1-pro"]
     // #then should have low and high variants
     expect(pro.variants).toBeTruthy()
     expect(pro.variants.low).toBeTruthy()

src/cli/config-manager/antigravity-provider-configuration.ts

@@ -4,10 +4,10 @@
  * IMPORTANT: Model names MUST use `antigravity-` prefix for stability.
  *
  * Since opencode-antigravity-auth v1.3.0, models use a variant system:
- * - `antigravity-gemini-3-pro` with variants: low, high
+ * - `antigravity-gemini-3.1-pro` with variants: low, high
  * - `antigravity-gemini-3-flash` with variants: minimal, low, medium, high
  *
- * Legacy tier-suffixed names (e.g., `antigravity-gemini-3-pro-high`) still work
+ * Legacy tier-suffixed names (e.g., `antigravity-gemini-3.1-pro-high`) still work
  * but variants are the recommended approach.
  *
  * @see https://github.com/NoeFabris/opencode-antigravity-auth#models
@@ -16,7 +16,7 @@ export const ANTIGRAVITY_PROVIDER_CONFIG = {
   google: {
     name: "Google",
     models: {
-      "antigravity-gemini-3-pro": {
+      "antigravity-gemini-3.1-pro": {
         name: "Gemini 3 Pro (Antigravity)",
         limit: { context: 1048576, output: 65535 },
         modalities: { input: ["text", "image", "pdf"], output: ["text"] },

src/cli/config-manager/bun-install.ts

@@ -1,4 +1,5 @@
 import { getConfigDir } from "./config-context"
+import { spawnWithWindowsHide } from "../../shared/spawn-with-windows-hide"
 
 const BUN_INSTALL_TIMEOUT_SECONDS = 60
 const BUN_INSTALL_TIMEOUT_MS = BUN_INSTALL_TIMEOUT_SECONDS * 1000
@@ -16,7 +17,7 @@ export async function runBunInstall(): Promise<boolean> {
 
 export async function runBunInstallWithDetails(): Promise<BunInstallResult> {
   try {
-    const proc = Bun.spawn(["bun", "install"], {
+    const proc = spawnWithWindowsHide(["bun", "install"], {
       cwd: getConfigDir(),
       stdout: "inherit",
       stderr: "inherit",

src/cli/config-manager/opencode-binary.ts

@@ -1,4 +1,5 @@
 import type { OpenCodeBinaryType } from "../../shared/opencode-config-dir-types"
+import { spawnWithWindowsHide } from "../../shared/spawn-with-windows-hide"
 import { initConfigContext } from "./config-context"
 
 const OPENCODE_BINARIES = ["opencode", "opencode-desktop"] as const
@@ -11,7 +12,7 @@ interface OpenCodeBinaryResult {
 async function findOpenCodeBinaryWithVersion(): Promise<OpenCodeBinaryResult | null> {
   for (const binary of OPENCODE_BINARIES) {
     try {
-      const proc = Bun.spawn([binary, "--version"], {
+      const proc = spawnWithWindowsHide([binary, "--version"], {
         stdout: "pipe",
         stderr: "pipe",
       })

src/cli/doctor/checks/dependencies.ts

@@ -3,6 +3,7 @@ import { createRequire } from "node:module"
 import { dirname, join } from "node:path"
 
 import type { DependencyInfo } from "../types"
+import { spawnWithWindowsHide } from "../../../shared/spawn-with-windows-hide"
 
 async function checkBinaryExists(binary: string): Promise<{ exists: boolean; path: string | null }> {
   try {
@@ -18,7 +19,7 @@ async function checkBinaryExists(binary: string): Promise<{ exists: boolean; pat
 
 async function getBinaryVersion(binary: string): Promise<string | null> {
   try {
-    const proc = Bun.spawn([binary, "--version"], { stdout: "pipe", stderr: "pipe" })
+    const proc = spawnWithWindowsHide([binary, "--version"], { stdout: "pipe", stderr: "pipe" })
     const output = await new Response(proc.stdout).text()
     await proc.exited
     if (proc.exitCode === 0) {
@@ -140,4 +141,3 @@ export async function checkCommentChecker(): Promise<DependencyInfo> {
     path: resolvedPath,
   }
 }
-

src/cli/doctor/checks/model-resolution.test.ts

@@ -26,7 +26,7 @@ describe("model-resolution check", () => {
       // then: Should have category entries
       const visual = info.categories.find((c) => c.name === "visual-engineering")
       expect(visual).toBeDefined()
-      expect(visual!.requirement.fallbackChain[0]?.model).toBe("gemini-3-pro")
+      expect(visual!.requirement.fallbackChain[0]?.model).toBe("gemini-3.1-pro")
       expect(visual!.requirement.fallbackChain[0]?.providers).toContain("google")
     })
   })

src/cli/doctor/checks/system-binary.ts

@@ -1,6 +1,7 @@
 import { existsSync } from "node:fs"
 import { homedir } from "node:os"
 import { join } from "node:path"
+import { spawnWithWindowsHide } from "../../../shared/spawn-with-windows-hide"
 
 import { OPENCODE_BINARIES } from "../constants"
 
@@ -110,7 +111,7 @@ export async function getOpenCodeVersion(
 ): Promise<string | null> {
   try {
     const command = buildVersionCommand(binaryPath, platform)
-    const processResult = Bun.spawn(command, { stdout: "pipe", stderr: "pipe" })
+    const processResult = spawnWithWindowsHide(command, { stdout: "pipe", stderr: "pipe" })
     const output = await new Response(processResult.stdout).text()
     await processResult.exited
 

src/cli/doctor/checks/tools-gh.ts

@@ -1,3 +1,5 @@
+import { spawnWithWindowsHide } from "../../../shared/spawn-with-windows-hide"
+
 export interface GhCliInfo {
   installed: boolean
   version: string | null
@@ -19,7 +21,7 @@ async function checkBinaryExists(binary: string): Promise<{ exists: boolean; pat
 
 async function getGhVersion(): Promise<string | null> {
   try {
-    const processResult = Bun.spawn(["gh", "--version"], { stdout: "pipe", stderr: "pipe" })
+    const processResult = spawnWithWindowsHide(["gh", "--version"], { stdout: "pipe", stderr: "pipe" })
     const output = await new Response(processResult.stdout).text()
     await processResult.exited
     if (processResult.exitCode !== 0) return null
@@ -38,7 +40,7 @@ async function getGhAuthStatus(): Promise<{
   error: string | null
 }> {
   try {
-    const processResult = Bun.spawn(["gh", "auth", "status"], {
+    const processResult = spawnWithWindowsHide(["gh", "auth", "status"], {
       stdout: "pipe",
       stderr: "pipe",
       env: { ...process.env, GH_NO_UPDATE_NOTIFIER: "1" },

src/cli/model-fallback-requirements.ts

@@ -24,7 +24,7 @@ export const CLI_AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
   oracle: {
     fallbackChain: [
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
     ],
   },
@@ -59,7 +59,7 @@ export const CLI_AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
       { providers: ["kimi-for-coding"], model: "k2p5" },
       { providers: ["opencode"], model: "kimi-k2.5-free" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro" },
     ],
   },
   metis: {
@@ -68,14 +68,14 @@ export const CLI_AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
       { providers: ["kimi-for-coding"], model: "k2p5" },
       { providers: ["opencode"], model: "kimi-k2.5-free" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
     ],
   },
   momus: {
     fallbackChain: [
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "medium" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
     ],
   },
   atlas: {
@@ -84,7 +84,7 @@ export const CLI_AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
       { providers: ["opencode"], model: "kimi-k2.5-free" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-sonnet-4-5" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro" },
     ],
   },
 }
@@ -92,7 +92,7 @@ export const CLI_AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
 export const CLI_CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
   "visual-engineering": {
     fallbackChain: [
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
       { providers: ["zai-coding-plan"], model: "glm-5" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
       { providers: ["kimi-for-coding"], model: "k2p5" },
@@ -101,7 +101,7 @@ export const CLI_CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> =
   ultrabrain: {
     fallbackChain: [
       { providers: ["openai", "opencode"], model: "gpt-5.3-codex", variant: "xhigh" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
     ],
   },
@@ -109,17 +109,17 @@ export const CLI_CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> =
     fallbackChain: [
       { providers: ["openai", "opencode"], model: "gpt-5.3-codex", variant: "medium" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
     ],
     requiresModel: "gpt-5.3-codex",
   },
   artistry: {
     fallbackChain: [
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2" },
     ],
-    requiresModel: "gemini-3-pro",
+    requiresModel: "gemini-3.1-pro",
   },
   quick: {
     fallbackChain: [
@@ -139,7 +139,7 @@ export const CLI_CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> =
     fallbackChain: [
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro" },
     ],
   },
   writing: {

src/cli/provider-model-id-transform.test.ts

@@ -40,16 +40,16 @@ describe("transformModelForProvider", () => {
 			expect(result).toBe("claude-haiku-4.5")
 		})
 
-		test("transforms gemini-3-pro to gemini-3-pro-preview", () => {
-			// #given github-copilot provider and gemini-3-pro model
+		test("transforms gemini-3.1-pro to gemini-3.1-pro-preview", () => {
+			// #given github-copilot provider and gemini-3.1-pro model
 			const provider = "github-copilot"
-			const model = "gemini-3-pro"
+			const model = "gemini-3.1-pro"
 
 			// #when transformModelForProvider is called
 			const result = transformModelForProvider(provider, model)
 
-			// #then should transform to gemini-3-pro-preview
-			expect(result).toBe("gemini-3-pro-preview")
+			// #then should transform to gemini-3.1-pro-preview
+			expect(result).toBe("gemini-3.1-pro-preview")
 		})
 
 		test("transforms gemini-3-flash to gemini-3-flash-preview", () => {
@@ -64,16 +64,16 @@ describe("transformModelForProvider", () => {
 			expect(result).toBe("gemini-3-flash-preview")
 		})
 
-		test("prevents double transformation of gemini-3-pro-preview", () => {
-			// #given github-copilot provider and gemini-3-pro-preview model (already transformed)
+		test("prevents double transformation of gemini-3.1-pro-preview", () => {
+			// #given github-copilot provider and gemini-3.1-pro-preview model (already transformed)
 			const provider = "github-copilot"
-			const model = "gemini-3-pro-preview"
+			const model = "gemini-3.1-pro-preview"
 
 			// #when transformModelForProvider is called
 			const result = transformModelForProvider(provider, model)
 
-			// #then should NOT become gemini-3-pro-preview-preview
-			expect(result).toBe("gemini-3-pro-preview")
+			// #then should NOT become gemini-3.1-pro-preview-preview
+			expect(result).toBe("gemini-3.1-pro-preview")
 		})
 
 		test("prevents double transformation of gemini-3-flash-preview", () => {
@@ -102,16 +102,16 @@ describe("transformModelForProvider", () => {
 			expect(result).toBe("gemini-3-flash-preview")
 		})
 
-		test("transforms gemini-3-pro to gemini-3-pro-preview", () => {
-			// #given google provider and gemini-3-pro model
+		test("transforms gemini-3.1-pro to gemini-3.1-pro-preview", () => {
+			// #given google provider and gemini-3.1-pro model
 			const provider = "google"
-			const model = "gemini-3-pro"
+			const model = "gemini-3.1-pro"
 
 			// #when transformModelForProvider is called
 			const result = transformModelForProvider(provider, model)
 
-			// #then should transform to gemini-3-pro-preview
-			expect(result).toBe("gemini-3-pro-preview")
+			// #then should transform to gemini-3.1-pro-preview
+			expect(result).toBe("gemini-3.1-pro-preview")
 		})
 
 		test("passes through other gemini models unchanged", () => {
@@ -138,16 +138,16 @@ describe("transformModelForProvider", () => {
 			expect(result).toBe("gemini-3-flash-preview")
 		})
 
-		test("prevents double transformation of gemini-3-pro-preview", () => {
-			// #given google provider and gemini-3-pro-preview model (already transformed)
+		test("prevents double transformation of gemini-3.1-pro-preview", () => {
+			// #given google provider and gemini-3.1-pro-preview model (already transformed)
 			const provider = "google"
-			const model = "gemini-3-pro-preview"
+			const model = "gemini-3.1-pro-preview"
 
 			// #when transformModelForProvider is called
 			const result = transformModelForProvider(provider, model)
 
-			// #then should NOT become gemini-3-pro-preview-preview
-			expect(result).toBe("gemini-3-pro-preview")
+			// #then should NOT become gemini-3.1-pro-preview-preview
+			expect(result).toBe("gemini-3.1-pro-preview")
 		})
 
 		test("does not transform claude models for google provider", () => {

src/cli/run/integration.test.ts

@@ -3,6 +3,7 @@ import type { RunResult } from "./types"
 import { createJsonOutputManager } from "./json-output"
 import { resolveSession } from "./session-resolver"
 import { executeOnCompleteHook } from "./on-complete-hook"
+import * as spawnWithWindowsHideModule from "../../shared/spawn-with-windows-hide"
 import type { OpencodeClient } from "./types"
 import * as originalSdk from "@opencode-ai/sdk"
 import * as originalPortUtils from "../../shared/port-utils"
@@ -147,7 +148,7 @@ describe("integration: --session-id", () => {
     const result = resolveSession({ client: mockClient, sessionId, directory: "/test" })
 
     // then
-    await expect(result).rejects.toThrow(`Session not found: ${sessionId}`)
+    expect(result).rejects.toThrow(`Session not found: ${sessionId}`)
     expect(mockClient.session.get).toHaveBeenCalledWith({
       path: { id: sessionId },
       query: { directory: "/test" },
@@ -161,10 +162,13 @@ describe("integration: --on-complete", () => {
 
   beforeEach(() => {
     spyOn(console, "error").mockImplementation(() => {})
-    spawnSpy = spyOn(Bun, "spawn").mockReturnValue({
+    spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockReturnValue({
       exited: Promise.resolve(0),
       exitCode: 0,
-    } as unknown as ReturnType<typeof Bun.spawn>)
+      stdout: undefined,
+      stderr: undefined,
+      kill: () => {},
+    } satisfies ReturnType<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>)
   })
 
   afterEach(() => {
@@ -186,7 +190,7 @@ describe("integration: --on-complete", () => {
 
     // then
     expect(spawnSpy).toHaveBeenCalledTimes(1)
-    const [_, options] = spawnSpy.mock.calls[0] as Parameters<typeof Bun.spawn>
+    const [_, options] = spawnSpy.mock.calls[0] as Parameters<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>
     expect(options?.env?.SESSION_ID).toBe("session-123")
     expect(options?.env?.EXIT_CODE).toBe("0")
     expect(options?.env?.DURATION_MS).toBe("5000")
@@ -208,10 +212,13 @@ describe("integration: option combinations", () => {
     spyOn(console, "error").mockImplementation(() => {})
     mockStdout = createMockWriteStream()
     mockStderr = createMockWriteStream()
-    spawnSpy = spyOn(Bun, "spawn").mockReturnValue({
+    spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockReturnValue({
       exited: Promise.resolve(0),
       exitCode: 0,
-    } as unknown as ReturnType<typeof Bun.spawn>)
+      stdout: undefined,
+      stderr: undefined,
+      kill: () => {},
+    } satisfies ReturnType<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>)
   })
 
   afterEach(() => {
@@ -249,9 +256,9 @@ describe("integration: option combinations", () => {
     const emitted = mockStdout.writes[0]!
     expect(() => JSON.parse(emitted)).not.toThrow()
     expect(spawnSpy).toHaveBeenCalledTimes(1)
-    const [args] = spawnSpy.mock.calls[0] as Parameters<typeof Bun.spawn>
+    const [args] = spawnSpy.mock.calls[0] as Parameters<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>
     expect(args).toEqual(["sh", "-c", "echo done"])
-    const [_, options] = spawnSpy.mock.calls[0] as Parameters<typeof Bun.spawn>
+    const [_, options] = spawnSpy.mock.calls[0] as Parameters<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>
     expect(options?.env?.SESSION_ID).toBe("session-123")
     expect(options?.env?.EXIT_CODE).toBe("0")
     expect(options?.env?.DURATION_MS).toBe("5000")

src/cli/run/on-complete-hook.test.ts

@@ -1,4 +1,5 @@
 import { describe, it, expect, spyOn, beforeEach, afterEach } from "bun:test"
+import * as spawnWithWindowsHideModule from "../../shared/spawn-with-windows-hide"
 import { executeOnCompleteHook } from "./on-complete-hook"
 
 describe("executeOnCompleteHook", () => {
@@ -6,7 +7,10 @@ describe("executeOnCompleteHook", () => {
     return {
       exited: Promise.resolve(exitCode),
       exitCode,
-    } as unknown as ReturnType<typeof Bun.spawn>
+      stdout: undefined,
+      stderr: undefined,
+      kill: () => {},
+    } satisfies ReturnType<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>
   }
 
   let consoleErrorSpy: ReturnType<typeof spyOn<typeof console, "error">>
@@ -21,7 +25,7 @@ describe("executeOnCompleteHook", () => {
 
   it("executes command with correct env vars", async () => {
     // given
-    const spawnSpy = spyOn(Bun, "spawn").mockReturnValue(createProc(0))
+    const spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockReturnValue(createProc(0))
 
     try {
       // when
@@ -35,7 +39,7 @@ describe("executeOnCompleteHook", () => {
 
       // then
       expect(spawnSpy).toHaveBeenCalledTimes(1)
-      const [args, options] = spawnSpy.mock.calls[0] as Parameters<typeof Bun.spawn>
+      const [args, options] = spawnSpy.mock.calls[0] as Parameters<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>
 
       expect(args).toEqual(["sh", "-c", "echo test"])
       expect(options?.env?.SESSION_ID).toBe("session-123")
@@ -51,7 +55,7 @@ describe("executeOnCompleteHook", () => {
 
   it("env var values are strings", async () => {
     // given
-    const spawnSpy = spyOn(Bun, "spawn").mockReturnValue(createProc(0))
+    const spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockReturnValue(createProc(0))
 
     try {
       // when
@@ -64,7 +68,7 @@ describe("executeOnCompleteHook", () => {
       })
 
       // then
-      const [_, options] = spawnSpy.mock.calls[0] as Parameters<typeof Bun.spawn>
+      const [_, options] = spawnSpy.mock.calls[0] as Parameters<typeof spawnWithWindowsHideModule.spawnWithWindowsHide>
 
       expect(options?.env?.EXIT_CODE).toBe("1")
       expect(options?.env?.EXIT_CODE).toBeTypeOf("string")
@@ -79,7 +83,7 @@ describe("executeOnCompleteHook", () => {
 
   it("empty command string is no-op", async () => {
     // given
-    const spawnSpy = spyOn(Bun, "spawn").mockReturnValue(createProc(0))
+    const spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockReturnValue(createProc(0))
 
     try {
       // when
@@ -100,7 +104,7 @@ describe("executeOnCompleteHook", () => {
 
   it("whitespace-only command is no-op", async () => {
     // given
-    const spawnSpy = spyOn(Bun, "spawn").mockReturnValue(createProc(0))
+    const spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockReturnValue(createProc(0))
 
     try {
       // when
@@ -121,11 +125,11 @@ describe("executeOnCompleteHook", () => {
 
   it("command failure logs warning but does not throw", async () => {
     // given
-    const spawnSpy = spyOn(Bun, "spawn").mockReturnValue(createProc(1))
+    const spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockReturnValue(createProc(1))
 
     try {
       // when
-      await expect(
+      expect(
         executeOnCompleteHook({
           command: "false",
           sessionId: "session-123",
@@ -149,13 +153,13 @@ describe("executeOnCompleteHook", () => {
   it("spawn error logs warning but does not throw", async () => {
     // given
     const spawnError = new Error("Command not found")
-    const spawnSpy = spyOn(Bun, "spawn").mockImplementation(() => {
+    const spawnSpy = spyOn(spawnWithWindowsHideModule, "spawnWithWindowsHide").mockImplementation(() => {
       throw spawnError
     })
 
     try {
       // when
-      await expect(
+      expect(
         executeOnCompleteHook({
           command: "nonexistent-command",
           sessionId: "session-123",

src/cli/run/on-complete-hook.ts

@@ -1,4 +1,5 @@
 import pc from "picocolors"
+import { spawnWithWindowsHide } from "../../shared/spawn-with-windows-hide"
 
 export async function executeOnCompleteHook(options: {
   command: string
@@ -17,7 +18,7 @@ export async function executeOnCompleteHook(options: {
   console.error(pc.dim(`Running on-complete hook: ${trimmedCommand}`))
 
   try {
-    const proc = Bun.spawn(["sh", "-c", trimmedCommand], {
+    const proc = spawnWithWindowsHide(["sh", "-c", trimmedCommand], {
       env: {
         ...process.env,
         SESSION_ID: sessionId,

src/cli/run/opencode-binary-resolver.ts

@@ -1,4 +1,5 @@
 import { delimiter, dirname, join } from "node:path"
+import { spawnWithWindowsHide } from "../../shared/spawn-with-windows-hide"
 
 const OPENCODE_COMMANDS = ["opencode", "opencode-desktop"] as const
 const WINDOWS_SUFFIXES = ["", ".exe", ".cmd", ".bat", ".ps1"] as const
@@ -41,7 +42,7 @@ export function collectCandidateBinaryPaths(
 
 export async function canExecuteBinary(binaryPath: string): Promise<boolean> {
   try {
-    const proc = Bun.spawn([binaryPath, "--version"], {
+    const proc = spawnWithWindowsHide([binaryPath, "--version"], {
       stdout: "pipe",
       stderr: "pipe",
     })

src/config/schema/oh-my-opencode-config.ts

@@ -27,7 +27,7 @@ export const OhMyOpenCodeConfigSchema = z.object({
   /** Default agent name for `oh-my-opencode run` (env: OPENCODE_DEFAULT_AGENT) */
   default_run_agent: z.string().optional(),
   disabled_mcps: z.array(AnyMcpNameSchema).optional(),
-  disabled_agents: z.array(BuiltinAgentNameSchema).optional(),
+  disabled_agents: z.array(z.string()).optional(),
   disabled_skills: z.array(BuiltinSkillNameSchema).optional(),
   disabled_hooks: z.array(z.string()).optional(),
   disabled_commands: z.array(BuiltinCommandNameSchema).optional(),

src/features/background-agent/concurrency.test.ts

@@ -34,7 +34,7 @@ describe("ConcurrencyManager.getConcurrencyLimit", () => {
   test("should return provider limit even when modelConcurrency exists but doesn't match", () => {
     // given
     const config: BackgroundTaskConfig = {
-      modelConcurrency: { "google/gemini-3-pro": 5 },
+      modelConcurrency: { "google/gemini-3.1-pro": 5 },
       providerConcurrency: { anthropic: 3 }
     }
     const manager = new ConcurrencyManager(config)
@@ -95,7 +95,7 @@ describe("ConcurrencyManager.getConcurrencyLimit", () => {
     // when
     const modelLimit = manager.getConcurrencyLimit("anthropic/claude-sonnet-4-6")
     const providerLimit = manager.getConcurrencyLimit("anthropic/claude-opus-4-6")
-    const defaultLimit = manager.getConcurrencyLimit("google/gemini-3-pro")
+    const defaultLimit = manager.getConcurrencyLimit("google/gemini-3.1-pro")
 
     // then
     expect(modelLimit).toBe(10)

src/features/background-agent/manager.test.ts

@@ -2987,6 +2987,28 @@ describe("BackgroundManager.handleEvent - session.deleted cascade", () => {
     manager.shutdown()
     resetToastManager()
   })
+
+  test("should clean pending notifications for deleted sessions", () => {
+    //#given
+    const manager = createBackgroundManager()
+    const sessionID = "session-pending-notifications"
+
+    manager.queuePendingNotification(sessionID, "<system-reminder>queued</system-reminder>")
+    expect(getPendingNotifications(manager).get(sessionID)).toEqual([
+      "<system-reminder>queued</system-reminder>",
+    ])
+
+    //#when
+    manager.handleEvent({
+      type: "session.deleted",
+      properties: { info: { id: sessionID } },
+    })
+
+    //#then
+    expect(getPendingNotifications(manager).has(sessionID)).toBe(false)
+
+    manager.shutdown()
+  })
 })
 
 describe("BackgroundManager.handleEvent - session.error", () => {

src/features/background-agent/manager.ts

@@ -830,6 +830,8 @@ export class BackgroundManager {
         tasksToCancel.set(descendant.id, descendant)
       }
 
+      this.pendingNotifications.delete(sessionID)
+
       if (tasksToCancel.size === 0) return
 
       for (const task of tasksToCancel.values()) {
@@ -866,6 +868,13 @@ export class BackgroundManager {
           subagentSessions.delete(task.sessionID)
         }
       }
+
+      for (const task of tasksToCancel.values()) {
+        if (task.parentSessionID) {
+          this.pendingNotifications.delete(task.parentSessionID)
+        }
+      }
+
       SessionCategoryRegistry.remove(sessionID)
     }
 

src/features/task-toast-manager/manager.test.ts

@@ -162,7 +162,7 @@ describe("TaskToastManager", () => {
         description: "Task with category default model",
         agent: "sisyphus-junior",
         isBackground: false,
-        modelInfo: { model: "google/gemini-3-pro", type: "category-default" as const },
+        modelInfo: { model: "google/gemini-3.1-pro", type: "category-default" as const },
       }
 
       // when - addTask is called

src/hooks/interactive-bash-session/interactive-bash-session-tracker.ts

@@ -6,6 +6,7 @@ import {
 import { OMO_SESSION_PREFIX, buildSessionReminderMessage } from "./constants";
 import type { InteractiveBashSessionState } from "./types";
 import { subagentSessions } from "../../features/claude-code-session-state";
+import { spawnWithWindowsHide } from "../../shared/spawn-with-windows-hide";
 
 type AbortSession = (args: { path: { id: string } }) => Promise<unknown>
 
@@ -19,7 +20,7 @@ async function killAllTrackedSessions(
 ): Promise<void> {
   for (const sessionName of state.tmuxSessions) {
     try {
-      const proc = Bun.spawn(["tmux", "kill-session", "-t", sessionName], {
+      const proc = spawnWithWindowsHide(["tmux", "kill-session", "-t", sessionName], {
         stdout: "ignore",
         stderr: "ignore",
       })

src/hooks/interactive-bash-session/state-manager.ts

@@ -1,6 +1,7 @@
 import type { InteractiveBashSessionState } from "./types";
 import { loadInteractiveBashSessionState } from "./storage";
 import { OMO_SESSION_PREFIX } from "./constants";
+import { spawnWithWindowsHide } from "../../shared/spawn-with-windows-hide";
 
 export function getOrCreateState(sessionID: string, sessionStates: Map<string, InteractiveBashSessionState>): InteractiveBashSessionState {
   if (!sessionStates.has(sessionID)) {
@@ -24,7 +25,7 @@ export async function killAllTrackedSessions(
 ): Promise<void> {
   for (const sessionName of state.tmuxSessions) {
     try {
-      const proc = Bun.spawn(["tmux", "kill-session", "-t", sessionName], {
+      const proc = spawnWithWindowsHide(["tmux", "kill-session", "-t", sessionName], {
         stdout: "ignore",
         stderr: "ignore",
       });

src/hooks/ralph-loop/completion-promise-detector.test.ts

@@ -0,0 +1,111 @@
+/// <reference types="bun-types" />
+import { describe, expect, test } from "bun:test"
+import type { PluginInput } from "@opencode-ai/plugin"
+import { detectCompletionInSessionMessages } from "./completion-promise-detector"
+
+type SessionMessage = {
+  info?: { role?: string }
+  parts?: Array<{ type: string; text?: string }>
+}
+
+function createPluginInput(messages: SessionMessage[]): PluginInput {
+  const pluginInput = {
+    client: { session: {} } as PluginInput["client"],
+    project: {} as PluginInput["project"],
+    directory: "/tmp",
+    worktree: "/tmp",
+    serverUrl: new URL("http://localhost"),
+    $: {} as PluginInput["$"],
+  } as PluginInput
+
+  pluginInput.client.session.messages =
+    (async () => ({ data: messages })) as unknown as PluginInput["client"]["session"]["messages"]
+
+  return pluginInput
+}
+
+describe("detectCompletionInSessionMessages", () => {
+  describe("#given session with prior DONE and new messages", () => {
+    test("#when sinceMessageIndex excludes prior DONE #then should NOT detect completion", async () => {
+      // #given
+      const messages: SessionMessage[] = [
+        {
+          info: { role: "assistant" },
+          parts: [{ type: "text", text: "Old completion <promise>DONE</promise>" }],
+        },
+        {
+          info: { role: "assistant" },
+          parts: [{ type: "text", text: "Working on the new task" }],
+        },
+      ]
+      const ctx = createPluginInput(messages)
+
+      // #when
+      const detected = await detectCompletionInSessionMessages(ctx, {
+        sessionID: "session-123",
+        promise: "DONE",
+        apiTimeoutMs: 1000,
+        directory: "/tmp",
+        sinceMessageIndex: 1,
+      })
+
+      // #then
+      expect(detected).toBe(false)
+    })
+
+    test("#when sinceMessageIndex includes current DONE #then should detect completion", async () => {
+      // #given
+      const messages: SessionMessage[] = [
+        {
+          info: { role: "assistant" },
+          parts: [{ type: "text", text: "Old completion <promise>DONE</promise>" }],
+        },
+        {
+          info: { role: "assistant" },
+          parts: [{ type: "text", text: "Current completion <promise>DONE</promise>" }],
+        },
+      ]
+      const ctx = createPluginInput(messages)
+
+      // #when
+      const detected = await detectCompletionInSessionMessages(ctx, {
+        sessionID: "session-123",
+        promise: "DONE",
+        apiTimeoutMs: 1000,
+        directory: "/tmp",
+        sinceMessageIndex: 1,
+      })
+
+      // #then
+      expect(detected).toBe(true)
+    })
+  })
+
+  describe("#given no sinceMessageIndex (backward compat)", () => {
+    test("#then should scan all messages", async () => {
+      // #given
+      const messages: SessionMessage[] = [
+        {
+          info: { role: "assistant" },
+          parts: [{ type: "text", text: "Old completion <promise>DONE</promise>" }],
+        },
+        {
+          info: { role: "assistant" },
+          parts: [{ type: "text", text: "No completion in latest message" }],
+        },
+      ]
+      const ctx = createPluginInput(messages)
+
+      // #when
+      const detected = await detectCompletionInSessionMessages(ctx, {
+        sessionID: "session-123",
+        promise: "DONE",
+        apiTimeoutMs: 1000,
+        directory: "/tmp",
+      })
+
+      // #then
+      expect(detected).toBe(true)
+    })
+  })
+})

src/hooks/ralph-loop/completion-promise-detector.ts

@@ -52,6 +52,7 @@ export async function detectCompletionInSessionMessages(
 		promise: string
 		apiTimeoutMs: number
 		directory: string
+		sinceMessageIndex?: number
 	},
 ): Promise<boolean> {
 	try {
@@ -75,12 +76,17 @@ export async function detectCompletionInSessionMessages(
 				? responseData
 				: []
 
-		const assistantMessages = (messageArray as OpenCodeSessionMessage[]).filter((msg) => msg.info?.role === "assistant")
+		const scopedMessages =
+			typeof options.sinceMessageIndex === "number" && options.sinceMessageIndex >= 0 && options.sinceMessageIndex < messageArray.length
+				? messageArray.slice(options.sinceMessageIndex)
+				: messageArray
+
+		const assistantMessages = (scopedMessages as OpenCodeSessionMessage[]).filter((msg) => msg.info?.role === "assistant")
 		if (assistantMessages.length === 0) return false
 
 		const pattern = buildPromisePattern(options.promise)
-		const recentAssistants = assistantMessages.slice(-3)
-		for (const assistant of recentAssistants) {
+		for (let index = assistantMessages.length - 1; index >= 0; index -= 1) {
+			const assistant = assistantMessages[index]
 			if (!assistant.parts) continue
 
 			let responseText = ""

src/hooks/ralph-loop/index.test.ts

@@ -494,6 +494,7 @@ describe("ralph-loop", () => {
         config: {
           enabled: true,
           default_max_iterations: 200,
+          default_strategy: "continue",
         },
       })
 
@@ -602,7 +603,7 @@ describe("ralph-loop", () => {
       expect(hook.getState()).toBeNull()
 
       // then - messages API was called with correct session ID
-      expect(messagesCalls.length).toBe(1)
+      expect(messagesCalls.length).toBe(2)
       expect(messagesCalls[0].sessionID).toBe("session-123")
     })
 
@@ -632,7 +633,7 @@ describe("ralph-loop", () => {
       expect(hook.getState()).toBeNull()
 
       // then - messages API was called with correct session ID
-      expect(messagesCalls.length).toBe(1)
+      expect(messagesCalls.length).toBe(2)
       expect(messagesCalls[0].sessionID).toBe("session-123")
     })
 
@@ -708,6 +709,57 @@ describe("ralph-loop", () => {
       expect(promptCalls[0].text).toContain("<promise>CALCULATOR_DONE</promise>")
     })
 
+    test("should skip concurrent idle events for same session when handler is in flight", async () => {
+      // given - active loop with delayed prompt injection
+      let releasePromptAsync: (() => void) | undefined
+      const promptAsyncBlocked = new Promise<void>((resolve) => {
+        releasePromptAsync = resolve
+      })
+      let firstPromptStartedResolve: (() => void) | undefined
+      const firstPromptStarted = new Promise<void>((resolve) => {
+        firstPromptStartedResolve = resolve
+      })
+
+      const mockInput = createMockPluginInput() as {
+        client: {
+          session: {
+            promptAsync: (opts: { path: { id: string }; body: { parts: Array<{ type: string; text: string }> } }) => Promise<unknown>
+          }
+        }
+      }
+
+      const originalPromptAsync = mockInput.client.session.promptAsync
+      let promptAsyncCalls = 0
+      mockInput.client.session.promptAsync = async (opts) => {
+        promptAsyncCalls += 1
+        if (promptAsyncCalls === 1) {
+          firstPromptStartedResolve?.()
+        }
+        await promptAsyncBlocked
+        return originalPromptAsync(opts)
+      }
+
+      const hook = createRalphLoopHook(mockInput as Parameters<typeof createRalphLoopHook>[0])
+      hook.startLoop("session-123", "Build feature", { maxIterations: 10 })
+
+      // when - second idle arrives while first idle processing is still in flight
+      const firstIdle = hook.event({
+        event: { type: "session.idle", properties: { sessionID: "session-123" } },
+      })
+      await firstPromptStarted
+      const secondIdle = hook.event({
+        event: { type: "session.idle", properties: { sessionID: "session-123" } },
+      })
+
+      releasePromptAsync?.()
+      await Promise.all([firstIdle, secondIdle])
+
+      // then - only one continuation should be injected
+      expect(promptAsyncCalls).toBe(1)
+      expect(promptCalls.length).toBe(1)
+      expect(hook.getState()?.iteration).toBe(2)
+    })
+
     test("should clear loop state on user abort (MessageAbortedError)", async () => {
       // given - active loop
       const hook = createRalphLoopHook(createMockPluginInput())
@@ -782,8 +834,8 @@ describe("ralph-loop", () => {
       expect(hook.getState()).toBeNull()
     })
 
-    test("should NOT detect completion if promise is older than last 3 assistant messages", async () => {
-      // given - promise appears in an assistant message older than last 3
+    test("should detect completion even when promise is older than previous narrow window", async () => {
+      // given - promise appears in an older assistant message with additional assistant output after it
       mockSessionMessages = [
         { info: { role: "user" }, parts: [{ type: "text", text: "Start task" }] },
         { info: { role: "assistant" }, parts: [{ type: "text", text: "Promise early <promise>DONE</promise>" }] },
@@ -801,9 +853,40 @@ describe("ralph-loop", () => {
         event: { type: "session.idle", properties: { sessionID: "session-123" } },
       })
 
-      // then - loop should continue (promise is older than last 3 assistant messages)
-      expect(promptCalls.length).toBe(1)
-      expect(hook.getState()?.iteration).toBe(2)
+      // then - loop should complete because all assistant messages are scanned
+      expect(promptCalls.length).toBe(0)
+      expect(toastCalls.some((t) => t.title === "Ralph Loop Complete!")).toBe(true)
+      expect(hook.getState()).toBeNull()
+    })
+
+    test("should detect completion when many assistant messages are emitted after promise", async () => {
+      // given - completion promise followed by long assistant output sequence
+      mockSessionMessages = [
+        { info: { role: "user" }, parts: [{ type: "text", text: "Start task" }] },
+        { info: { role: "assistant" }, parts: [{ type: "text", text: "Done now <promise>DONE</promise>" }] },
+      ]
+
+      for (let index = 1; index <= 25; index += 1) {
+        mockSessionMessages.push({
+          info: { role: "assistant" },
+          parts: [{ type: "text", text: `Post-completion assistant output ${index}` }],
+        })
+      }
+
+      const hook = createRalphLoopHook(createMockPluginInput(), {
+        getTranscriptPath: () => join(TEST_DIR, "nonexistent.jsonl"),
+      })
+      hook.startLoop("session-123", "Build something", { completionPromise: "DONE" })
+
+      // when - session goes idle
+      await hook.event({
+        event: { type: "session.idle", properties: { sessionID: "session-123" } },
+      })
+
+      // then - loop should complete despite large trailing output
+      expect(promptCalls.length).toBe(0)
+      expect(toastCalls.some((t) => t.title === "Ralph Loop Complete!")).toBe(true)
+      expect(hook.getState()).toBeNull()
     })
 
     test("should allow starting new loop while previous loop is active (different session)", async () => {
@@ -992,7 +1075,7 @@ Original task: Build something`
       expect(promptCalls.length).toBe(0)
       expect(hook.getState()).toBeNull()
       // API should NOT be called since transcript found completion
-      expect(messagesCalls.length).toBe(0)
+      expect(messagesCalls.length).toBe(1)
     })
 
     test("should show ultrawork completion toast", async () => {

src/hooks/ralph-loop/loop-state-controller.ts

@@ -23,6 +23,7 @@ export function createLoopStateController(options: {
 			loopOptions?: {
 				maxIterations?: number
 				completionPromise?: string
+				messageCountAtStart?: number
 				ultrawork?: boolean
 				strategy?: "reset" | "continue"
 			},
@@ -34,6 +35,7 @@ export function createLoopStateController(options: {
 					loopOptions?.maxIterations ??
 					config?.default_max_iterations ??
 					DEFAULT_MAX_ITERATIONS,
+				message_count_at_start: loopOptions?.messageCountAtStart,
 				completion_promise:
 					loopOptions?.completionPromise ??
 					DEFAULT_COMPLETION_PROMISE,
@@ -93,5 +95,19 @@ export function createLoopStateController(options: {
 
 			return state
 		},
+
+		setMessageCountAtStart(sessionID: string, messageCountAtStart: number): RalphLoopState | null {
+			const state = readState(directory, stateDir)
+			if (!state || state.session_id !== sessionID) {
+				return null
+			}
+
+			state.message_count_at_start = messageCountAtStart
+			if (!writeState(directory, state, stateDir)) {
+				return null
+			}
+
+			return state
+		},
 	}
 }

src/hooks/ralph-loop/ralph-loop-event-handler.ts

@@ -25,6 +25,8 @@ export function createRalphLoopEventHandler(
 	ctx: PluginInput,
 	options: RalphLoopEventHandlerOptions,
 ) {
+	const inFlightSessions = new Set<string>()
+
 	return async ({ event }: { event: { type: string; properties?: unknown } }): Promise<void> => {
 		const props = event.properties as Record<string, unknown> | undefined
 
@@ -32,115 +34,128 @@ export function createRalphLoopEventHandler(
 			const sessionID = props?.sessionID as string | undefined
 			if (!sessionID) return
 
-			if (options.sessionRecovery.isRecovering(sessionID)) {
-				log(`[${HOOK_NAME}] Skipped: in recovery`, { sessionID })
+			if (inFlightSessions.has(sessionID)) {
+				log(`[${HOOK_NAME}] Skipped: handler in flight`, { sessionID })
 				return
 			}
 
-			const state = options.loopState.getState()
-			if (!state || !state.active) {
-				return
-			}
-
-			if (state.session_id && state.session_id !== sessionID) {
-				if (options.checkSessionExists) {
-					try {
-						const exists = await options.checkSessionExists(state.session_id)
-						if (!exists) {
-							options.loopState.clear()
-							log(`[${HOOK_NAME}] Cleared orphaned state from deleted session`, {
-								orphanedSessionId: state.session_id,
-								currentSessionId: sessionID,
-							})
-							return
-						}
-					} catch (err) {
-						log(`[${HOOK_NAME}] Failed to check session existence`, {
-							sessionId: state.session_id,
-							error: String(err),
-						})
-					}
-				}
-				return
-			}
-
-			const transcriptPath = options.getTranscriptPath(sessionID)
-			const completionViaTranscript = detectCompletionInTranscript(transcriptPath, state.completion_promise)
-			const completionViaApi = completionViaTranscript
-				? false
-				: await detectCompletionInSessionMessages(ctx, {
-					sessionID,
-					promise: state.completion_promise,
-					apiTimeoutMs: options.apiTimeoutMs,
-					directory: options.directory,
-				})
-
-			if (completionViaTranscript || completionViaApi) {
-				log(`[${HOOK_NAME}] Completion detected!`, {
-					sessionID,
-					iteration: state.iteration,
-					promise: state.completion_promise,
-					detectedVia: completionViaTranscript
-						? "transcript_file"
-						: "session_messages_api",
-				})
-				options.loopState.clear()
-
-				const title = state.ultrawork ? "ULTRAWORK LOOP COMPLETE!" : "Ralph Loop Complete!"
-				const message = state.ultrawork ? `JUST ULW ULW! Task completed after ${state.iteration} iteration(s)` : `Task completed after ${state.iteration} iteration(s)`
-				await ctx.client.tui?.showToast?.({ body: { title, message, variant: "success", duration: 5000 } }).catch(() => {})
-				return
-			}
-
-			if (state.iteration >= state.max_iterations) {
-				log(`[${HOOK_NAME}] Max iterations reached`, {
-					sessionID,
-					iteration: state.iteration,
-					max: state.max_iterations,
-				})
-				options.loopState.clear()
-
-				await ctx.client.tui?.showToast?.({
-					body: { title: "Ralph Loop Stopped", message: `Max iterations (${state.max_iterations}) reached without completion`, variant: "warning", duration: 5000 },
-					}).catch(() => {})
-				return
-			}
-
-			const newState = options.loopState.incrementIteration()
-			if (!newState) {
-				log(`[${HOOK_NAME}] Failed to increment iteration`, { sessionID })
-				return
-			}
-
-			log(`[${HOOK_NAME}] Continuing loop`, {
-				sessionID,
-				iteration: newState.iteration,
-				max: newState.max_iterations,
-			})
-
-			await ctx.client.tui?.showToast?.({
-				body: {
-					title: "Ralph Loop",
-					message: `Iteration ${newState.iteration}/${newState.max_iterations}`,
-					variant: "info",
-					duration: 2000,
-				},
-				}).catch(() => {})
+			inFlightSessions.add(sessionID)
 
 			try {
-				await continueIteration(ctx, newState, {
-					previousSessionID: sessionID,
-					directory: options.directory,
-					apiTimeoutMs: options.apiTimeoutMs,
-					loopState: options.loopState,
-				})
-			} catch (err) {
-				log(`[${HOOK_NAME}] Failed to inject continuation`, {
+
+				if (options.sessionRecovery.isRecovering(sessionID)) {
+					log(`[${HOOK_NAME}] Skipped: in recovery`, { sessionID })
+					return
+				}
+
+				const state = options.loopState.getState()
+				if (!state || !state.active) {
+					return
+				}
+
+				if (state.session_id && state.session_id !== sessionID) {
+					if (options.checkSessionExists) {
+						try {
+							const exists = await options.checkSessionExists(state.session_id)
+							if (!exists) {
+								options.loopState.clear()
+								log(`[${HOOK_NAME}] Cleared orphaned state from deleted session`, {
+									orphanedSessionId: state.session_id,
+									currentSessionId: sessionID,
+								})
+								return
+							}
+						} catch (err) {
+							log(`[${HOOK_NAME}] Failed to check session existence`, {
+								sessionId: state.session_id,
+								error: String(err),
+							})
+						}
+					}
+					return
+				}
+
+				const transcriptPath = options.getTranscriptPath(sessionID)
+				const completionViaTranscript = detectCompletionInTranscript(transcriptPath, state.completion_promise)
+				const completionViaApi = completionViaTranscript
+					? false
+					: await detectCompletionInSessionMessages(ctx, {
+						sessionID,
+						promise: state.completion_promise,
+						apiTimeoutMs: options.apiTimeoutMs,
+						directory: options.directory,
+						sinceMessageIndex: state.message_count_at_start,
+					})
+
+				if (completionViaTranscript || completionViaApi) {
+					log(`[${HOOK_NAME}] Completion detected!`, {
+						sessionID,
+						iteration: state.iteration,
+						promise: state.completion_promise,
+						detectedVia: completionViaTranscript
+							? "transcript_file"
+							: "session_messages_api",
+					})
+					options.loopState.clear()
+
+					const title = state.ultrawork ? "ULTRAWORK LOOP COMPLETE!" : "Ralph Loop Complete!"
+					const message = state.ultrawork ? `JUST ULW ULW! Task completed after ${state.iteration} iteration(s)` : `Task completed after ${state.iteration} iteration(s)`
+					await ctx.client.tui?.showToast?.({ body: { title, message, variant: "success", duration: 5000 } }).catch(() => {})
+					return
+				}
+
+				if (state.iteration >= state.max_iterations) {
+					log(`[${HOOK_NAME}] Max iterations reached`, {
+						sessionID,
+						iteration: state.iteration,
+						max: state.max_iterations,
+					})
+					options.loopState.clear()
+
+					await ctx.client.tui?.showToast?.({
+						body: { title: "Ralph Loop Stopped", message: `Max iterations (${state.max_iterations}) reached without completion`, variant: "warning", duration: 5000 },
+						}).catch(() => {})
+					return
+				}
+
+				const newState = options.loopState.incrementIteration()
+				if (!newState) {
+					log(`[${HOOK_NAME}] Failed to increment iteration`, { sessionID })
+					return
+				}
+
+				log(`[${HOOK_NAME}] Continuing loop`, {
 					sessionID,
-					error: String(err),
+					iteration: newState.iteration,
+					max: newState.max_iterations,
 				})
+
+				await ctx.client.tui?.showToast?.({
+					body: {
+						title: "Ralph Loop",
+						message: `Iteration ${newState.iteration}/${newState.max_iterations}`,
+						variant: "info",
+						duration: 2000,
+					},
+					}).catch(() => {})
+
+				try {
+					await continueIteration(ctx, newState, {
+						previousSessionID: sessionID,
+						directory: options.directory,
+						apiTimeoutMs: options.apiTimeoutMs,
+						loopState: options.loopState,
+					})
+				} catch (err) {
+					log(`[${HOOK_NAME}] Failed to inject continuation`, {
+						sessionID,
+						error: String(err),
+					})
+				}
+				return
+			} finally {
+				inFlightSessions.delete(sessionID)
 			}
-			return
 		}
 
 		if (event.type === "session.deleted") {

src/hooks/ralph-loop/ralph-loop-hook.ts

@@ -13,6 +13,7 @@ export interface RalphLoopHook {
     options?: {
       maxIterations?: number
       completionPromise?: string
+      messageCountAtStart?: number
       ultrawork?: boolean
       strategy?: "reset" | "continue"
     }
@@ -23,6 +24,19 @@ export interface RalphLoopHook {
 
 const DEFAULT_API_TIMEOUT = 5000 as const
 
+function getMessageCountFromResponse(messagesResponse: unknown): number {
+  if (Array.isArray(messagesResponse)) {
+    return messagesResponse.length
+  }
+
+  if (typeof messagesResponse === "object" && messagesResponse !== null && "data" in messagesResponse) {
+    const data = (messagesResponse as { data?: unknown }).data
+    return Array.isArray(data) ? data.length : 0
+  }
+
+  return 0
+}
+
 export function createRalphLoopHook(
   ctx: PluginInput,
   options?: RalphLoopOptions
@@ -51,7 +65,25 @@ export function createRalphLoopHook(
 
 	return {
 		event,
-		startLoop: loopState.startLoop,
+		startLoop: (sessionID, prompt, loopOptions): boolean => {
+			const startSuccess = loopState.startLoop(sessionID, prompt, loopOptions)
+			if (!startSuccess || typeof loopOptions?.messageCountAtStart === "number") {
+				return startSuccess
+			}
+
+			ctx.client.session
+				.messages({
+					path: { id: sessionID },
+					query: { directory: ctx.directory },
+				})
+				.then((messagesResponse: unknown) => {
+					const messageCountAtStart = getMessageCountFromResponse(messagesResponse)
+					loopState.setMessageCountAtStart(sessionID, messageCountAtStart)
+				})
+				.catch(() => {})
+
+			return startSuccess
+		},
 		cancelLoop: loopState.cancelLoop,
 		getState: loopState.getState as () => RalphLoopState | null,
 	}

src/hooks/ralph-loop/reset-strategy-race-condition.test.ts

@@ -36,7 +36,7 @@ async function waitUntil(condition: () => boolean): Promise<void> {
 }
 
 describe("ralph-loop reset strategy race condition", () => {
-  test("should continue iteration when old session idle arrives before TUI switch completes", async () => {
+  test("should skip duplicate idle while reset iteration handling is in flight", async () => {
     // given - reset strategy loop with blocked TUI session switch
     const promptCalls: Array<{ sessionID: string; text: string }> = []
     const createSessionCalls: Array<{ parentID?: string }> = []
@@ -85,7 +85,7 @@ describe("ralph-loop reset strategy race condition", () => {
           },
         },
       },
-    } as Parameters<typeof createRalphLoopHook>[0])
+    } as unknown as Parameters<typeof createRalphLoopHook>[0])
 
     hook.startLoop("session-old", "Build feature", { strategy: "reset" })
 
@@ -100,14 +100,12 @@ describe("ralph-loop reset strategy race condition", () => {
       event: { type: "session.idle", properties: { sessionID: "session-old" } },
     })
 
-    await waitUntil(() => selectSessionCalls > 1)
-
     selectSessionDeferred.resolve()
     await Promise.all([firstIdleEvent, secondIdleEvent])
 
-    // then - second idle should not be skipped during reset transition
-    expect(createSessionCalls.length).toBe(2)
-    expect(promptCalls.length).toBe(2)
-    expect(hook.getState()?.iteration).toBe(3)
+    // then - duplicate idle should be skipped to prevent concurrent continuation injection
+    expect(createSessionCalls.length).toBe(1)
+    expect(promptCalls.length).toBe(1)
+    expect(hook.getState()?.iteration).toBe(2)
   })
 })

src/hooks/ralph-loop/storage.ts

@@ -44,6 +44,12 @@ export function readState(directory: string, customPath?: string): RalphLoopStat
       active: isActive,
       iteration: iterationNum,
       max_iterations: Number(data.max_iterations) || DEFAULT_MAX_ITERATIONS,
+      message_count_at_start:
+        typeof data.message_count_at_start === "number"
+          ? data.message_count_at_start
+          : typeof data.message_count_at_start === "string" && data.message_count_at_start.trim() !== ""
+            ? Number(data.message_count_at_start)
+            : undefined,
       completion_promise: stripQuotes(data.completion_promise) || DEFAULT_COMPLETION_PROMISE,
       started_at: stripQuotes(data.started_at) || new Date().toISOString(),
       prompt: body.trim(),
@@ -72,13 +78,17 @@ export function writeState(
     const sessionIdLine = state.session_id ? `session_id: "${state.session_id}"\n` : ""
     const ultraworkLine = state.ultrawork !== undefined ? `ultrawork: ${state.ultrawork}\n` : ""
     const strategyLine = state.strategy ? `strategy: "${state.strategy}"\n` : ""
+    const messageCountAtStartLine =
+      typeof state.message_count_at_start === "number"
+        ? `message_count_at_start: ${state.message_count_at_start}\n`
+        : ""
     const content = `---
 active: ${state.active}
 iteration: ${state.iteration}
 max_iterations: ${state.max_iterations}
 completion_promise: "${state.completion_promise}"
 started_at: "${state.started_at}"
-${sessionIdLine}${ultraworkLine}${strategyLine}---
+${sessionIdLine}${ultraworkLine}${strategyLine}${messageCountAtStartLine}---
 ${state.prompt}
 `
 

src/hooks/ralph-loop/types.ts

@@ -4,6 +4,7 @@ export interface RalphLoopState {
   active: boolean
   iteration: number
   max_iterations: number
+  message_count_at_start?: number
   completion_promise: string
   started_at: string
   prompt: string

src/hooks/runtime-fallback/index.test.ts

@@ -125,7 +125,7 @@ describe("runtime-fallback", () => {
       await hook.event({
         event: {
           type: "session.created",
-          properties: { info: { id: sessionID, model: "google/gemini-3-pro" } },
+          properties: { info: { id: sessionID, model: "google/gemini-3.1-pro" } },
         },
       })
 
@@ -1841,7 +1841,7 @@ describe("runtime-fallback", () => {
     test("should apply fallback model on next chat.message after error", async () => {
       const hook = createRuntimeFallbackHook(createMockPluginInput(), {
         config: createMockConfig({ notify_on_fallback: false }),
-        pluginConfig: createMockPluginConfigWithCategoryFallback(["openai/gpt-5.2", "google/gemini-3-pro"]),
+        pluginConfig: createMockPluginConfigWithCategoryFallback(["openai/gpt-5.2", "google/gemini-3.1-pro"]),
       })
       const sessionID = "test-session-switch"
       SessionCategoryRegistry.register(sessionID, "test")
@@ -1916,7 +1916,7 @@ describe("runtime-fallback", () => {
       const input = createMockPluginInput()
       const hook = createRuntimeFallbackHook(input, {
         config: createMockConfig({ notify_on_fallback: false }),
-        pluginConfig: createMockPluginConfigWithAgentFallback("oracle", ["openai/gpt-5.2", "google/gemini-3-pro"]),
+        pluginConfig: createMockPluginConfigWithAgentFallback("oracle", ["openai/gpt-5.2", "google/gemini-3.1-pro"]),
       })
       const sessionID = "test-agent-fallback"
 

src/hooks/stop-continuation-guard/hook.ts

@@ -1,4 +1,5 @@
 import type { PluginInput } from "@opencode-ai/plugin"
+import type { BackgroundManager } from "../../features/background-agent"
 
 import {
   clearContinuationMarker,
@@ -8,6 +9,11 @@ import { log } from "../../shared/logger"
 
 const HOOK_NAME = "stop-continuation-guard"
 
+type StopContinuationBackgroundManager = Pick<
+  BackgroundManager,
+  "getAllDescendantTasks" | "cancelTask"
+>
+
 export interface StopContinuationGuard {
   event: (input: { event: { type: string; properties?: unknown } }) => Promise<void>
   "chat.message": (input: { sessionID?: string }) => Promise<void>
@@ -17,7 +23,10 @@ export interface StopContinuationGuard {
 }
 
 export function createStopContinuationGuardHook(
-  ctx: PluginInput
+  ctx: PluginInput,
+  options?: {
+    backgroundManager?: StopContinuationBackgroundManager
+  }
 ): StopContinuationGuard {
   const stoppedSessions = new Set<string>()
 
@@ -25,6 +34,38 @@ export function createStopContinuationGuardHook(
     stoppedSessions.add(sessionID)
     setContinuationMarkerSource(ctx.directory, sessionID, "stop", "stopped", "continuation stopped")
     log(`[${HOOK_NAME}] Continuation stopped for session`, { sessionID })
+
+    const backgroundManager = options?.backgroundManager
+    if (!backgroundManager) {
+      return
+    }
+
+    const cancellableTasks = backgroundManager
+      .getAllDescendantTasks(sessionID)
+      .filter((task) => task.status === "running" || task.status === "pending")
+
+    if (cancellableTasks.length === 0) {
+      return
+    }
+
+    void Promise.allSettled(
+      cancellableTasks.map(async (task) => {
+        await backgroundManager.cancelTask(task.id, {
+          source: "stop-continuation",
+          reason: "Continuation stopped via /stop-continuation",
+          abortSession: task.status === "running",
+          skipNotification: true,
+        })
+      })
+    ).then((results) => {
+      const cancelledCount = results.filter((result) => result.status === "fulfilled").length
+      const failedCount = results.length - cancelledCount
+      log(`[${HOOK_NAME}] Cancelled background tasks for stopped session`, {
+        sessionID,
+        cancelledCount,
+        failedCount,
+      })
+    })
   }
 
   const isStopped = (sessionID: string): boolean => {

src/hooks/stop-continuation-guard/index.test.ts

@@ -2,9 +2,15 @@ import { afterEach, describe, expect, test } from "bun:test"
 import { mkdtempSync, rmSync } from "node:fs"
 import { join } from "node:path"
 import { tmpdir } from "node:os"
+import type { BackgroundManager, BackgroundTask } from "../../features/background-agent"
 import { readContinuationMarker } from "../../features/run-continuation-state"
 import { createStopContinuationGuardHook } from "./index"
 
+type CancelCall = {
+  taskId: string
+  options?: Parameters<BackgroundManager["cancelTask"]>[1]
+}
+
 describe("stop-continuation-guard", () => {
   const tempDirs: string[] = []
 
@@ -34,6 +40,33 @@ describe("stop-continuation-guard", () => {
     } as any
   }
 
+  function createBackgroundTask(status: BackgroundTask["status"], id: string): BackgroundTask {
+    return {
+      id,
+      status,
+      description: `${id} description`,
+      parentSessionID: "parent-session",
+      parentMessageID: "parent-message",
+      prompt: "prompt",
+      agent: "sisyphus-junior",
+    }
+  }
+
+  function createMockBackgroundManager(tasks: BackgroundTask[], cancelCalls: CancelCall[]): Pick<BackgroundManager, "getAllDescendantTasks" | "cancelTask"> {
+    return {
+      getAllDescendantTasks: () => tasks,
+      cancelTask: async (taskId: string, options?: Parameters<BackgroundManager["cancelTask"]>[1]) => {
+        cancelCalls.push({ taskId, options })
+        return true
+      },
+    }
+  }
+
+  async function flushMicrotasks(): Promise<void> {
+    await Promise.resolve()
+    await Promise.resolve()
+  }
+
   test("should mark session as stopped", () => {
     // given - a guard hook with no stopped sessions
     const input = createMockPluginInput()
@@ -166,4 +199,31 @@ describe("stop-continuation-guard", () => {
     // then - should not throw and stopped session remains stopped
     expect(guard.isStopped("some-session")).toBe(true)
   })
+
+  test("should cancel only running and pending background tasks on stop", async () => {
+    // given - a background manager with mixed task statuses
+    const cancelCalls: CancelCall[] = []
+    const backgroundManager = createMockBackgroundManager(
+      [
+        createBackgroundTask("running", "task-running"),
+        createBackgroundTask("pending", "task-pending"),
+        createBackgroundTask("completed", "task-completed"),
+      ],
+      cancelCalls,
+    )
+    const guard = createStopContinuationGuardHook(createMockPluginInput(), {
+      backgroundManager,
+    })
+
+    // when - stop continuation is triggered
+    guard.stop("test-session-bg")
+    await flushMicrotasks()
+
+    // then - only running and pending tasks are cancelled
+    expect(cancelCalls).toHaveLength(2)
+    expect(cancelCalls[0]?.taskId).toBe("task-running")
+    expect(cancelCalls[0]?.options?.abortSession).toBe(true)
+    expect(cancelCalls[1]?.taskId).toBe("task-pending")
+    expect(cancelCalls[1]?.options?.abortSession).toBe(false)
+  })
 })

src/hooks/think-mode/index.test.ts

@@ -109,7 +109,7 @@ describe("createThinkModeHook", () => {
     const input = createHookInput({
       sessionID,
       providerID: "google",
-      modelID: "gemini-3-pro",
+      modelID: "gemini-3.1-pro",
     })
     const output = createHookOutput("Please solve this directly")
 

src/hooks/think-mode/switcher.test.ts

@@ -49,8 +49,8 @@ describe("think-mode switcher", () => {
 
       it("should handle Gemini preview variants", () => {
         // given Gemini preview model IDs
-        expect(getHighVariant("gemini-3-pro")).toBe(
-          "gemini-3-pro-high"
+        expect(getHighVariant("gemini-3.1-pro")).toBe(
+          "gemini-3-1-pro-high"
         )
         expect(getHighVariant("gemini-3-flash")).toBe(
           "gemini-3-flash-high"
@@ -61,7 +61,7 @@ describe("think-mode switcher", () => {
         // given model IDs that are already high variants
         expect(getHighVariant("claude-opus-4-6-high")).toBeNull()
         expect(getHighVariant("gpt-5-2-high")).toBeNull()
-        expect(getHighVariant("gemini-3-pro-high")).toBeNull()
+        expect(getHighVariant("gemini-3-1-pro-high")).toBeNull()
       })
 
       it("should return null for unknown models", () => {
@@ -77,7 +77,7 @@ describe("think-mode switcher", () => {
       // given model IDs with -high suffix
       expect(isAlreadyHighVariant("claude-opus-4-6-high")).toBe(true)
       expect(isAlreadyHighVariant("gpt-5-2-high")).toBe(true)
-      expect(isAlreadyHighVariant("gemini-3-pro-high")).toBe(true)
+      expect(isAlreadyHighVariant("gemini-3.1-pro-high")).toBe(true)
     })
 
     it("should detect -high suffix after normalization", () => {
@@ -90,7 +90,7 @@ describe("think-mode switcher", () => {
       expect(isAlreadyHighVariant("claude-opus-4-6")).toBe(false)
       expect(isAlreadyHighVariant("claude-opus-4.6")).toBe(false)
       expect(isAlreadyHighVariant("gpt-5.2")).toBe(false)
-      expect(isAlreadyHighVariant("gemini-3-pro")).toBe(false)
+      expect(isAlreadyHighVariant("gemini-3.1-pro")).toBe(false)
     })
 
     it("should return false for models with 'high' in name but not suffix", () => {
@@ -129,7 +129,7 @@ describe("think-mode switcher", () => {
         // given various custom prefixes
         expect(getHighVariant("azure/gpt-5")).toBe("azure/gpt-5-high")
         expect(getHighVariant("bedrock/claude-sonnet-4-6")).toBe("bedrock/claude-sonnet-4-6-high")
-        expect(getHighVariant("custom-llm/gemini-3-pro")).toBe("custom-llm/gemini-3-pro-high")
+        expect(getHighVariant("custom-llm/gemini-3.1-pro")).toBe("custom-llm/gemini-3-1-pro-high")
       })
 
       it("should return null for prefixed models without high variant mapping", () => {
@@ -150,7 +150,7 @@ describe("think-mode switcher", () => {
         // given prefixed model IDs with -high suffix
         expect(isAlreadyHighVariant("vertex_ai/claude-opus-4-6-high")).toBe(true)
         expect(isAlreadyHighVariant("openai/gpt-5-2-high")).toBe(true)
-        expect(isAlreadyHighVariant("custom/gemini-3-pro-high")).toBe(true)
+        expect(isAlreadyHighVariant("custom/gemini-3.1-pro-high")).toBe(true)
       })
 
       it("should return false for prefixed base models", () => {
@@ -167,4 +167,3 @@ describe("think-mode switcher", () => {
     })
 })
 })
-

src/hooks/think-mode/switcher.ts

@@ -62,8 +62,8 @@ const HIGH_VARIANT_MAP: Record<string, string> = {
   "claude-sonnet-4-6": "claude-sonnet-4-6-high",
   "claude-opus-4-6": "claude-opus-4-6-high",
    // Gemini
-   "gemini-3-pro": "gemini-3-pro-high",
-   "gemini-3-pro-low": "gemini-3-pro-high",
+   "gemini-3-1-pro": "gemini-3-1-pro-high",
+   "gemini-3-1-pro-low": "gemini-3-1-pro-high",
    "gemini-3-flash": "gemini-3-flash-high",
   // GPT-5
   "gpt-5": "gpt-5-high",
@@ -82,7 +82,7 @@ const HIGH_VARIANT_MAP: Record<string, string> = {
   "gpt-5-2-chat-latest": "gpt-5-2-chat-latest-high",
   "gpt-5-2-pro": "gpt-5-2-pro-high",
   // Antigravity (Google)
-  "antigravity-gemini-3-pro": "antigravity-gemini-3-pro-high",
+  "antigravity-gemini-3-1-pro": "antigravity-gemini-3-1-pro-high",
   "antigravity-gemini-3-flash": "antigravity-gemini-3-flash-high",
 }
 
@@ -114,4 +114,3 @@ export function isAlreadyHighVariant(modelID: string): boolean {
   return ALREADY_HIGH.has(base) || base.endsWith("-high")
 }
 
-

src/hooks/todo-continuation-enforcer/idle-event.ts

@@ -15,6 +15,7 @@ import {
   MAX_CONSECUTIVE_FAILURES,
 } from "./constants"
 import { isLastAssistantMessageAborted } from "./abort-detection"
+import { hasUnansweredQuestion } from "./pending-question-detection"
 import { getIncompleteCount } from "./todo"
 import type { MessageInfo, ResolvedMessageInfo, Todo } from "./types"
 import type { SessionStateStore } from "./session-state"
@@ -74,6 +75,10 @@ export async function handleSessionIdle(args: {
       log(`[${HOOK_NAME}] Skipped: last assistant message was aborted (API fallback)`, { sessionID })
       return
     }
+    if (hasUnansweredQuestion(messages)) {
+      log(`[${HOOK_NAME}] Skipped: pending question awaiting user response`, { sessionID })
+      return
+    }
   } catch (error) {
     log(`[${HOOK_NAME}] Messages fetch failed, continuing`, { sessionID, error: String(error) })
   }

src/hooks/todo-continuation-enforcer/pending-question-detection.test.ts

@@ -0,0 +1,100 @@
+/// <reference types="bun-types" />
+import { describe, expect, test } from "bun:test"
+
+import { hasUnansweredQuestion } from "./pending-question-detection"
+
+describe("hasUnansweredQuestion", () => {
+  test("given empty messages, returns false", () => {
+    expect(hasUnansweredQuestion([])).toBe(false)
+  })
+
+  test("given null-ish input, returns false", () => {
+    expect(hasUnansweredQuestion(undefined as never)).toBe(false)
+  })
+
+  test("given last assistant message with question tool_use, returns true", () => {
+    const messages = [
+      { info: { role: "user" } },
+      {
+        info: { role: "assistant" },
+        parts: [
+          { type: "tool_use", name: "question" },
+        ],
+      },
+    ]
+    expect(hasUnansweredQuestion(messages)).toBe(true)
+  })
+
+  test("given last assistant message with question tool-invocation, returns true", () => {
+    const messages = [
+      { info: { role: "user" } },
+      {
+        info: { role: "assistant" },
+        parts: [
+          { type: "tool-invocation", toolName: "question" },
+        ],
+      },
+    ]
+    expect(hasUnansweredQuestion(messages)).toBe(true)
+  })
+
+  test("given user message after question (answered), returns false", () => {
+    const messages = [
+      {
+        info: { role: "assistant" },
+        parts: [
+          { type: "tool_use", name: "question" },
+        ],
+      },
+      { info: { role: "user" } },
+    ]
+    expect(hasUnansweredQuestion(messages)).toBe(false)
+  })
+
+  test("given assistant message with non-question tool, returns false", () => {
+    const messages = [
+      { info: { role: "user" } },
+      {
+        info: { role: "assistant" },
+        parts: [
+          { type: "tool_use", name: "bash" },
+        ],
+      },
+    ]
+    expect(hasUnansweredQuestion(messages)).toBe(false)
+  })
+
+  test("given assistant message with no parts, returns false", () => {
+    const messages = [
+      { info: { role: "user" } },
+      { info: { role: "assistant" } },
+    ]
+    expect(hasUnansweredQuestion(messages)).toBe(false)
+  })
+
+  test("given role on message directly (not in info), returns true for question", () => {
+    const messages = [
+      { role: "user" },
+      {
+        role: "assistant",
+        parts: [
+          { type: "tool_use", name: "question" },
+        ],
+      },
+    ]
+    expect(hasUnansweredQuestion(messages)).toBe(true)
+  })
+
+  test("given mixed tools including question, returns true", () => {
+    const messages = [
+      {
+        info: { role: "assistant" },
+        parts: [
+          { type: "tool_use", name: "bash" },
+          { type: "tool_use", name: "question" },
+        ],
+      },
+    ]
+    expect(hasUnansweredQuestion(messages)).toBe(true)
+  })
+})

src/hooks/todo-continuation-enforcer/pending-question-detection.ts

@@ -0,0 +1,40 @@
+import { log } from "../../shared/logger"
+import { HOOK_NAME } from "./constants"
+
+interface MessagePart {
+  type: string
+  name?: string
+  toolName?: string
+}
+
+interface Message {
+  info?: { role?: string }
+  role?: string
+  parts?: MessagePart[]
+}
+
+export function hasUnansweredQuestion(messages: Message[]): boolean {
+  if (!messages || messages.length === 0) return false
+
+  for (let i = messages.length - 1; i >= 0; i--) {
+    const msg = messages[i]
+    const role = msg.info?.role ?? msg.role
+
+    if (role === "user") return false
+
+    if (role === "assistant" && msg.parts) {
+      const hasQuestion = msg.parts.some(
+        (part) =>
+          (part.type === "tool_use" || part.type === "tool-invocation") &&
+          (part.name === "question" || part.toolName === "question"),
+      )
+      if (hasQuestion) {
+        log(`[${HOOK_NAME}] Detected pending question tool in last assistant message`)
+        return true
+      }
+      return false
+    }
+  }
+
+  return false
+}

src/hooks/todo-continuation-enforcer/todo-continuation-enforcer.test.ts

@@ -297,6 +297,31 @@ describe("todo-continuation-enforcer", () => {
     expect(promptCalls).toHaveLength(0)
   })
 
+  test("should not inject when remaining todos are blocked or deleted", async () => {
+    // given - session where non-completed todos are only blocked/deleted
+    const sessionID = "main-blocked-deleted"
+    setMainSession(sessionID)
+
+    const mockInput = createMockPluginInput()
+    mockInput.client.session.todo = async () => ({ data: [
+      { id: "1", content: "Blocked task", status: "blocked", priority: "high" },
+      { id: "2", content: "Deleted task", status: "deleted", priority: "medium" },
+      { id: "3", content: "Done task", status: "completed", priority: "low" },
+    ]})
+
+    const hook = createTodoContinuationEnforcer(mockInput, {})
+
+    // when - session goes idle
+    await hook.handler({
+      event: { type: "session.idle", properties: { sessionID } },
+    })
+
+    await fakeTimers.advanceBy(3000)
+
+    // then - no continuation injected
+    expect(promptCalls).toHaveLength(0)
+  })
+
   test("should not inject when background tasks are running", async () => {
     // given - session with running background tasks
     const sessionID = "main-789"
@@ -1663,7 +1688,6 @@ describe("todo-continuation-enforcer", () => {
   test("should cancel all countdowns via cancelAllCountdowns", async () => {
     // given - multiple sessions with running countdowns
     const session1 = "main-cancel-all-1"
-    const session2 = "main-cancel-all-2"
     setMainSession(session1)
 
     const hook = createTodoContinuationEnforcer(createMockPluginInput(), {})

src/hooks/todo-continuation-enforcer/todo.ts

@@ -1,5 +1,11 @@
 import type { Todo } from "./types"
 
 export function getIncompleteCount(todos: Todo[]): number {
-  return todos.filter((todo) => todo.status !== "completed" && todo.status !== "cancelled").length
+  return todos.filter(
+    (todo) =>
+      todo.status !== "completed"
+      && todo.status !== "cancelled"
+      && todo.status !== "blocked"
+      && todo.status !== "deleted",
+  ).length
 }

src/plugin-config.test.ts

@@ -27,7 +27,7 @@ describe("mergeConfigs", () => {
             temperature: 0.3,
           },
           visual: {
-            model: "google/gemini-3-pro",
+            model: "google/gemini-3.1-pro",
           },
         },
       } as unknown as OhMyOpenCodeConfig;
@@ -41,7 +41,7 @@ describe("mergeConfigs", () => {
       // then quick should be preserved from base
       expect(result.categories?.quick?.model).toBe("anthropic/claude-haiku-4-5");
       // then visual should be added from override
-      expect(result.categories?.visual?.model).toBe("google/gemini-3-pro");
+      expect(result.categories?.visual?.model).toBe("google/gemini-3.1-pro");
     });
 
     it("should preserve base categories when override has no categories", () => {

src/plugin-handlers/agent-config-handler.ts

@@ -106,6 +106,15 @@ export async function applyAgentConfig(params: {
     ]),
   );
 
+  const disabledAgentNames = new Set(
+    (migratedDisabledAgents ?? []).map(a => a.toLowerCase())
+  );
+
+  const filterDisabledAgents = (agents: Record<string, unknown>) =>
+    Object.fromEntries(
+      Object.entries(agents).filter(([name]) => !disabledAgentNames.has(name.toLowerCase()))
+    );
+
   const isSisyphusEnabled = params.pluginConfig.sisyphus_agent?.disabled !== true;
   const builderEnabled =
     params.pluginConfig.sisyphus_agent?.default_builder_enabled ?? false;
@@ -194,9 +203,9 @@ export async function applyAgentConfig(params: {
       ...Object.fromEntries(
         Object.entries(builtinAgents).filter(([key]) => key !== "sisyphus"),
       ),
-      ...userAgents,
-      ...projectAgents,
-      ...pluginAgents,
+      ...filterDisabledAgents(userAgents),
+      ...filterDisabledAgents(projectAgents),
+      ...filterDisabledAgents(pluginAgents),
       ...filteredConfigAgents,
       build: { ...migratedBuild, mode: "subagent", hidden: true },
       ...(planDemoteConfig ? { plan: planDemoteConfig } : {}),
@@ -204,9 +213,9 @@ export async function applyAgentConfig(params: {
   } else {
     params.config.agent = {
       ...builtinAgents,
-      ...userAgents,
-      ...projectAgents,
-      ...pluginAgents,
+      ...filterDisabledAgents(userAgents),
+      ...filterDisabledAgents(projectAgents),
+      ...filterDisabledAgents(pluginAgents),
       ...configAgent,
     };
   }

src/plugin-handlers/config-handler.test.ts

@@ -570,7 +570,7 @@ describe("Prometheus category config resolution", () => {
 
     // then
     expect(config).toBeDefined()
-    expect(config?.model).toBe("google/gemini-3-pro")
+    expect(config?.model).toBe("google/gemini-3.1-pro")
   })
 
   test("user categories override default categories", () => {

src/plugin/hooks/create-continuation-hooks.ts

@@ -49,7 +49,10 @@ export function createContinuationHooks(args: {
     safeCreateHook(hookName, factory, { enabled: safeHookEnabled })
 
   const stopContinuationGuard = isHookEnabled("stop-continuation-guard")
-    ? safeHook("stop-continuation-guard", () => createStopContinuationGuardHook(ctx))
+    ? safeHook("stop-continuation-guard", () =>
+        createStopContinuationGuardHook(ctx, {
+          backgroundManager,
+        }))
     : null
 
   const compactionContextInjector = isHookEnabled("compaction-context-injector")

src/plugin/ultrawork-db-model-override.ts

@@ -23,8 +23,8 @@ function tryUpdateMessageModel(
   if (result.changes === 0) return false
   if (variant) {
     db.prepare(
-      `UPDATE message SET data = json_set(data, '$.variant', ?) WHERE id = ?`,
-    ).run(variant, messageId)
+      `UPDATE message SET data = json_set(data, '$.variant', ?, '$.thinking', ?) WHERE id = ?`,
+    ).run(variant, variant, messageId)
   }
   return true
 }

src/plugin/ultrawork-model-override.test.ts

@@ -279,6 +279,30 @@ describe("applyUltraworkModelOverrideOnMessage", () => {
     )
   })
 
+  test("should override keyword-detector variant with configured ultrawork variant on deferred path", () => {
+    //#given
+    const config = createConfig("sisyphus", {
+      model: "anthropic/claude-opus-4-6",
+      variant: "extended",
+    })
+    const output = createOutput("ultrawork do something", { messageId: "msg_123" })
+    output.message["variant"] = "max"
+    output.message["thinking"] = "max"
+    const tui = createMockTui()
+
+    //#when
+    applyUltraworkModelOverrideOnMessage(config, "sisyphus", output, tui)
+
+    //#then
+    expect(dbOverrideSpy).toHaveBeenCalledWith(
+      "msg_123",
+      { providerID: "anthropic", modelID: "claude-opus-4-6" },
+      "extended",
+    )
+    expect(output.message["variant"]).toBe("extended")
+    expect(output.message["thinking"]).toBe("extended")
+  })
+
   test("should NOT mutate output.message.model when message ID present", () => {
     //#given
     const sonnetModel = { providerID: "anthropic", modelID: "claude-sonnet-4-6" }

src/plugin/ultrawork-model-override.ts

@@ -114,10 +114,12 @@ export function applyUltraworkModelOverrideOnMessage(
   const override = resolveUltraworkOverride(pluginConfig, inputAgentName, output, sessionID)
   if (!override) return
 
+  if (override.variant) {
+    output.message["variant"] = override.variant
+    output.message["thinking"] = override.variant
+  }
+
   if (!override.providerID || !override.modelID) {
-    if (override.variant) {
-      output.message["variant"] = override.variant
-    }
     return
   }
 
@@ -131,10 +133,8 @@ export function applyUltraworkModelOverrideOnMessage(
   if (!messageId) {
     log("[ultrawork-model-override] No message ID found, falling back to direct mutation")
     output.message.model = targetModel
-    if (override.variant) {
-      output.message["variant"] = override.variant
-    }
     return
+
   }
 
   const fromModel = (output.message.model as { modelID?: string } | undefined)?.modelID ?? "unknown"

src/shared/migration.test.ts

@@ -774,7 +774,7 @@ describe("migrateAgentConfigToCategory", () => {
   test("migrates model to category when mapping exists", () => {
     // given: Config with a model that has a category mapping
     const config = {
-      model: "google/gemini-3-pro",
+      model: "google/gemini-3.1-pro",
       temperature: 0.5,
       top_p: 0.9,
     }
@@ -823,7 +823,7 @@ describe("migrateAgentConfigToCategory", () => {
   test("handles all mapped models correctly", () => {
     // given: Configs for each mapped model
     const configs = [
-      { model: "google/gemini-3-pro" },
+      { model: "google/gemini-3.1-pro" },
       { model: "google/gemini-3-flash" },
       { model: "openai/gpt-5.2" },
       { model: "anthropic/claude-haiku-4-5" },
@@ -893,7 +893,7 @@ describe("shouldDeleteAgentConfig", () => {
     // given: Config with fields matching category defaults
     const config = {
       category: "visual-engineering",
-      model: "google/gemini-3-pro",
+      model: "google/gemini-3.1-pro",
     }
 
     // when: Check if config should be deleted
@@ -1021,7 +1021,7 @@ describe("migrateConfigFile with backup", () => {
       agents: {
         "multimodal-looker": { model: "anthropic/claude-haiku-4-5" },
         oracle: { model: "openai/gpt-5.2" },
-        "my-custom-agent": { model: "google/gemini-3-pro" },
+        "my-custom-agent": { model: "google/gemini-3.1-pro" },
       },
     }
 
@@ -1037,7 +1037,7 @@ describe("migrateConfigFile with backup", () => {
     const agents = rawConfig.agents as Record<string, Record<string, unknown>>
     expect(agents["multimodal-looker"].model).toBe("anthropic/claude-haiku-4-5")
     expect(agents.oracle.model).toBe("openai/gpt-5.2")
-    expect(agents["my-custom-agent"].model).toBe("google/gemini-3-pro")
+    expect(agents["my-custom-agent"].model).toBe("google/gemini-3.1-pro")
   })
 
   test("preserves category setting when explicitly set", () => {

src/shared/migration/agent-category.ts

@@ -12,7 +12,7 @@
  * This map will be removed in a future major version once migration period ends.
  */
 export const MODEL_TO_CATEGORY_MAP: Record<string, string> = {
-  "google/gemini-3-pro": "visual-engineering",
+  "google/gemini-3.1-pro": "visual-engineering",
   "google/gemini-3-flash": "writing",
   "openai/gpt-5.2": "ultrabrain",
   "anthropic/claude-haiku-4-5": "quick",

src/shared/model-availability.test.ts

@@ -63,7 +63,7 @@ describe("fetchAvailableModels", () => {
     writeModelsCache({
       openai: { id: "openai", models: { "gpt-5.2": { id: "gpt-5.2" } } },
       anthropic: { id: "anthropic", models: { "claude-opus-4-6": { id: "claude-opus-4-6" } } },
-      google: { id: "google", models: { "gemini-3-pro": { id: "gemini-3-pro" } } },
+      google: { id: "google", models: { "gemini-3.1-pro": { id: "gemini-3.1-pro" } } },
     })
 
     const result = await fetchAvailableModels(undefined, {
@@ -74,7 +74,7 @@ describe("fetchAvailableModels", () => {
     expect(result.size).toBe(3)
     expect(result.has("openai/gpt-5.2")).toBe(true)
     expect(result.has("anthropic/claude-opus-4-6")).toBe(true)
-    expect(result.has("google/gemini-3-pro")).toBe(true)
+    expect(result.has("google/gemini-3.1-pro")).toBe(true)
   })
 
   it("#given connectedProviders unknown #when fetchAvailableModels called without options #then returns empty Set", async () => {
@@ -97,7 +97,7 @@ describe("fetchAvailableModels", () => {
         list: async () => ({
           data: [
             { id: "gpt-5.3-codex", provider: "openai" },
-            { id: "gemini-3-pro", provider: "google" },
+            { id: "gemini-3.1-pro", provider: "google" },
           ],
         }),
       },
@@ -107,7 +107,7 @@ describe("fetchAvailableModels", () => {
 
     expect(result).toBeInstanceOf(Set)
     expect(result.has("openai/gpt-5.3-codex")).toBe(true)
-    expect(result.has("google/gemini-3-pro")).toBe(false)
+    expect(result.has("google/gemini-3.1-pro")).toBe(false)
   })
 
   it("#given cache file not found #when fetchAvailableModels called with connectedProviders #then returns empty Set", async () => {
@@ -126,7 +126,7 @@ describe("fetchAvailableModels", () => {
         list: async () => ({
           data: [
             { id: "gpt-5.3-codex", provider: "openai" },
-            { id: "gemini-3-pro", provider: "google" },
+            { id: "gemini-3.1-pro", provider: "google" },
           ],
         }),
       },
@@ -136,7 +136,7 @@ describe("fetchAvailableModels", () => {
 
     expect(result).toBeInstanceOf(Set)
     expect(result.has("openai/gpt-5.3-codex")).toBe(true)
-    expect(result.has("google/gemini-3-pro")).toBe(true)
+    expect(result.has("google/gemini-3.1-pro")).toBe(true)
   })
 
   it("#given cache read twice #when second call made with same providers #then reads fresh each time", async () => {
@@ -515,7 +515,7 @@ describe("fetchAvailableModels with connected providers filtering", () => {
 		writeModelsCache({
 			openai: { models: { "gpt-5.2": { id: "gpt-5.2" } } },
 			anthropic: { models: { "claude-opus-4-6": { id: "claude-opus-4-6" } } },
-			google: { models: { "gemini-3-pro": { id: "gemini-3-pro" } } },
+			google: { models: { "gemini-3.1-pro": { id: "gemini-3.1-pro" } } },
 		})
 
 		const result = await fetchAvailableModels(undefined, {
@@ -525,7 +525,7 @@ describe("fetchAvailableModels with connected providers filtering", () => {
 		expect(result.size).toBe(1)
 		expect(result.has("anthropic/claude-opus-4-6")).toBe(true)
 		expect(result.has("openai/gpt-5.2")).toBe(false)
-		expect(result.has("google/gemini-3-pro")).toBe(false)
+		expect(result.has("google/gemini-3.1-pro")).toBe(false)
 	})
 
 	// given cache with multiple providers
@@ -535,7 +535,7 @@ describe("fetchAvailableModels with connected providers filtering", () => {
 		writeModelsCache({
 			openai: { models: { "gpt-5.2": { id: "gpt-5.2" } } },
 			anthropic: { models: { "claude-opus-4-6": { id: "claude-opus-4-6" } } },
-			google: { models: { "gemini-3-pro": { id: "gemini-3-pro" } } },
+			google: { models: { "gemini-3.1-pro": { id: "gemini-3.1-pro" } } },
 		})
 
 		const result = await fetchAvailableModels(undefined, {
@@ -544,7 +544,7 @@ describe("fetchAvailableModels with connected providers filtering", () => {
 
 		expect(result.size).toBe(2)
 		expect(result.has("anthropic/claude-opus-4-6")).toBe(true)
-		expect(result.has("google/gemini-3-pro")).toBe(true)
+		expect(result.has("google/gemini-3.1-pro")).toBe(true)
 		expect(result.has("openai/gpt-5.2")).toBe(false)
 	})
 
@@ -759,7 +759,7 @@ describe("fetchAvailableModels with provider-models cache (whitelist-filtered)",
 			models: {
 				opencode: ["big-pickle"],
 				anthropic: ["claude-opus-4-6"],
-				google: ["gemini-3-pro"]
+				google: ["gemini-3.1-pro"]
 			},
 			connected: ["opencode", "anthropic", "google"]
 		})
@@ -771,7 +771,7 @@ describe("fetchAvailableModels with provider-models cache (whitelist-filtered)",
 		expect(result.size).toBe(1)
 		expect(result.has("opencode/big-pickle")).toBe(true)
 		expect(result.has("anthropic/claude-opus-4-6")).toBe(false)
-		expect(result.has("google/gemini-3-pro")).toBe(false)
+		expect(result.has("google/gemini-3.1-pro")).toBe(false)
 	})
 
 	it("should handle object[] format with metadata (Ollama-style)", async () => {
@@ -953,7 +953,7 @@ describe("fallback model availability", () => {
 			{ providers: ["openai"], model: "gpt-5.2" },
 			{ providers: ["anthropic"], model: "claude-opus-4-6" },
 		]
-		const availableModels = new Set(["google/gemini-3-pro"])
+		const availableModels = new Set(["google/gemini-3.1-pro"])
 
 		// when
 		const result = resolveFirstAvailableFallback(fallbackChain, availableModels)

src/shared/model-requirements.test.ts

@@ -248,19 +248,19 @@ describe("CATEGORY_MODEL_REQUIREMENTS", () => {
     expect(primary.providers[0]).toBe("openai")
   })
 
-  test("visual-engineering has valid fallbackChain with gemini-3-pro high as primary", () => {
+  test("visual-engineering has valid fallbackChain with gemini-3.1-pro high as primary", () => {
     // given - visual-engineering category requirement
     const visualEngineering = CATEGORY_MODEL_REQUIREMENTS["visual-engineering"]
 
     // when - accessing visual-engineering requirement
-    // then - fallbackChain: gemini-3-pro(high) → glm-5 → opus-4-6(max)
+    // then - fallbackChain: gemini-3.1-pro(high) → glm-5 → opus-4-6(max)
     expect(visualEngineering).toBeDefined()
     expect(visualEngineering.fallbackChain).toBeArray()
     expect(visualEngineering.fallbackChain).toHaveLength(3)
 
     const primary = visualEngineering.fallbackChain[0]
     expect(primary.providers[0]).toBe("google")
-    expect(primary.model).toBe("gemini-3-pro")
+    expect(primary.model).toBe("gemini-3.1-pro")
     expect(primary.variant).toBe("high")
 
     const second = visualEngineering.fallbackChain[1]
@@ -319,39 +319,43 @@ describe("CATEGORY_MODEL_REQUIREMENTS", () => {
     expect(primary.providers).toEqual(["anthropic", "github-copilot", "opencode"])
   })
 
-  test("artistry has valid fallbackChain with gemini-3-pro as primary", () => {
+  test("artistry has valid fallbackChain with gemini-3.1-pro as primary", () => {
     // given - artistry category requirement
     const artistry = CATEGORY_MODEL_REQUIREMENTS["artistry"]
 
     // when - accessing artistry requirement
-    // then - fallbackChain exists with gemini-3-pro as first entry
+    // then - fallbackChain exists with gemini-3.1-pro as first entry
     expect(artistry).toBeDefined()
     expect(artistry.fallbackChain).toBeArray()
     expect(artistry.fallbackChain.length).toBeGreaterThan(0)
 
     const primary = artistry.fallbackChain[0]
-    expect(primary.model).toBe("gemini-3-pro")
+    expect(primary.model).toBe("gemini-3.1-pro")
     expect(primary.variant).toBe("high")
     expect(primary.providers[0]).toBe("google")
   })
 
-  test("writing has valid fallbackChain with gemini-3-flash as primary", () => {
+  test("writing has valid fallbackChain with kimi-k2.5-free as primary", () => {
     // given - writing category requirement
     const writing = CATEGORY_MODEL_REQUIREMENTS["writing"]
 
     // when - accessing writing requirement
-    // then - fallbackChain: gemini-3-flash → claude-sonnet-4-6
+    // then - fallbackChain: kimi-k2.5-free -> gemini-3-flash -> claude-sonnet-4-6
     expect(writing).toBeDefined()
     expect(writing.fallbackChain).toBeArray()
-    expect(writing.fallbackChain).toHaveLength(2)
+    expect(writing.fallbackChain).toHaveLength(3)
 
     const primary = writing.fallbackChain[0]
-    expect(primary.model).toBe("gemini-3-flash")
-    expect(primary.providers[0]).toBe("google")
+    expect(primary.model).toBe("kimi-k2.5-free")
+    expect(primary.providers[0]).toBe("opencode")
 
     const second = writing.fallbackChain[1]
-    expect(second.model).toBe("claude-sonnet-4-6")
-    expect(second.providers[0]).toBe("anthropic")
+    expect(second.model).toBe("gemini-3-flash")
+    expect(second.providers[0]).toBe("google")
+
+    const third = writing.fallbackChain[2]
+    expect(third.model).toBe("claude-sonnet-4-6")
+    expect(third.providers[0]).toBe("anthropic")
   })
 
   test("all 8 categories have valid fallbackChain arrays", () => {
@@ -489,12 +493,12 @@ describe("requiresModel field in categories", () => {
     expect(deep.requiresModel).toBe("gpt-5.3-codex")
   })
 
-  test("artistry category has requiresModel set to gemini-3-pro", () => {
+  test("artistry category has requiresModel set to gemini-3.1-pro", () => {
     // given
     const artistry = CATEGORY_MODEL_REQUIREMENTS["artistry"]
 
     // when / #then
-    expect(artistry.requiresModel).toBe("gemini-3-pro")
+    expect(artistry.requiresModel).toBe("gemini-3.1-pro")
   })
 })
 

src/shared/model-requirements.ts

@@ -32,7 +32,7 @@ export const AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
   oracle: {
     fallbackChain: [
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
     ],
   },
@@ -65,7 +65,7 @@ export const AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
       { providers: ["opencode"], model: "kimi-k2.5-free" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro" },
     ],
   },
   metis: {
@@ -73,14 +73,14 @@ export const AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
       { providers: ["opencode"], model: "kimi-k2.5-free" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
     ],
   },
   momus: {
     fallbackChain: [
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "medium" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
     ],
   },
   atlas: {
@@ -95,7 +95,7 @@ export const AGENT_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
 export const CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
   "visual-engineering": {
     fallbackChain: [
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { 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" },
     ],
@@ -103,7 +103,7 @@ export const CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
   ultrabrain: {
     fallbackChain: [
       { providers: ["openai", "opencode"], model: "gpt-5.3-codex", variant: "xhigh" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
     ],
   },
@@ -111,17 +111,17 @@ export const CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
     fallbackChain: [
       { providers: ["openai", "opencode"], model: "gpt-5.3-codex", variant: "medium" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
     ],
     requiresModel: "gpt-5.3-codex",
   },
   artistry: {
     fallbackChain: [
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro", variant: "high" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro", variant: "high" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2" },
     ],
-    requiresModel: "gemini-3-pro",
+    requiresModel: "gemini-3.1-pro",
   },
   quick: {
     fallbackChain: [
@@ -141,11 +141,12 @@ export const CATEGORY_MODEL_REQUIREMENTS: Record<string, ModelRequirement> = {
     fallbackChain: [
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6", variant: "max" },
       { providers: ["openai", "github-copilot", "opencode"], model: "gpt-5.2", variant: "high" },
-      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro" },
+      { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro" },
     ],
   },
   writing: {
     fallbackChain: [
+      { providers: ["opencode"], model: "kimi-k2.5-free" },
       { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-flash" },
       { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-sonnet-4-6" },
     ],

src/shared/model-resolver.test.ts

@@ -10,7 +10,7 @@ describe("resolveModel", () => {
       const input: ModelResolutionInput = {
         userModel: "anthropic/claude-opus-4-6",
         inheritedModel: "openai/gpt-5.2",
-        systemDefault: "google/gemini-3-pro",
+        systemDefault: "google/gemini-3.1-pro",
       }
 
       // when
@@ -25,7 +25,7 @@ describe("resolveModel", () => {
       const input: ModelResolutionInput = {
         userModel: undefined,
         inheritedModel: "openai/gpt-5.2",
-        systemDefault: "google/gemini-3-pro",
+        systemDefault: "google/gemini-3.1-pro",
       }
 
       // when
@@ -40,14 +40,14 @@ describe("resolveModel", () => {
       const input: ModelResolutionInput = {
         userModel: undefined,
         inheritedModel: undefined,
-        systemDefault: "google/gemini-3-pro",
+        systemDefault: "google/gemini-3.1-pro",
       }
 
       // when
       const result = resolveModel(input)
 
       // then
-      expect(result).toBe("google/gemini-3-pro")
+      expect(result).toBe("google/gemini-3.1-pro")
     })
   })
 
@@ -57,7 +57,7 @@ describe("resolveModel", () => {
       const input: ModelResolutionInput = {
         userModel: "",
         inheritedModel: "openai/gpt-5.2",
-        systemDefault: "google/gemini-3-pro",
+        systemDefault: "google/gemini-3.1-pro",
       }
 
       // when
@@ -72,14 +72,14 @@ describe("resolveModel", () => {
       const input: ModelResolutionInput = {
         userModel: "   ",
         inheritedModel: "",
-        systemDefault: "google/gemini-3-pro",
+        systemDefault: "google/gemini-3.1-pro",
       }
 
       // when
       const result = resolveModel(input)
 
       // then
-      expect(result).toBe("google/gemini-3-pro")
+      expect(result).toBe("google/gemini-3.1-pro")
     })
   })
 
@@ -89,7 +89,7 @@ describe("resolveModel", () => {
       const input: ModelResolutionInput = {
         userModel: "anthropic/claude-opus-4-6",
         inheritedModel: "openai/gpt-5.2",
-        systemDefault: "google/gemini-3-pro",
+        systemDefault: "google/gemini-3.1-pro",
       }
 
       // when
@@ -123,7 +123,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic", "github-copilot"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(["anthropic/claude-opus-4-6", "github-copilot/claude-opus-4-6-preview"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -141,7 +141,7 @@ describe("resolveModelWithFallback", () => {
         uiSelectedModel: "opencode/big-pickle",
         userModel: "anthropic/claude-opus-4-6",
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -158,7 +158,7 @@ describe("resolveModelWithFallback", () => {
         uiSelectedModel: "   ",
         userModel: "anthropic/claude-opus-4-6",
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -175,7 +175,7 @@ describe("resolveModelWithFallback", () => {
         uiSelectedModel: "",
         userModel: "anthropic/claude-opus-4-6",
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -195,7 +195,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic", "github-copilot"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(["anthropic/claude-opus-4-6", "github-copilot/claude-opus-4-6-preview"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -215,7 +215,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -234,7 +234,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -252,7 +252,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -271,7 +271,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic", "github-copilot", "opencode"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(["github-copilot/claude-opus-4-6-preview", "opencode/claude-opus-4-7"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -294,8 +294,8 @@ describe("resolveModelWithFallback", () => {
         fallbackChain: [
           { providers: ["openai", "anthropic", "google"], model: "gpt-5.2" },
         ],
-        availableModels: new Set(["openai/gpt-5.2", "anthropic/claude-opus-4-6", "google/gemini-3-pro"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        availableModels: new Set(["openai/gpt-5.2", "anthropic/claude-opus-4-6", "google/gemini-3.1-pro"]),
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -313,7 +313,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic", "opencode"], model: "gpt-5-nano" },
         ],
         availableModels: new Set(["opencode/gpt-5-nano"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -331,7 +331,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic", "github-copilot"], model: "claude-opus" },
         ],
         availableModels: new Set(["anthropic/claude-opus-4-6", "github-copilot/claude-opus-4-6-preview"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -346,7 +346,7 @@ describe("resolveModelWithFallback", () => {
       // given
       const input: ExtendedModelResolutionInput = {
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -361,7 +361,7 @@ describe("resolveModelWithFallback", () => {
       const input: ExtendedModelResolutionInput = {
         fallbackChain: [],
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -378,7 +378,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "CLAUDE-OPUS" },
         ],
         availableModels: new Set(["anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -397,7 +397,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "claude-sonnet-4-6" },
         ],
         availableModels: new Set(["opencode/glm-5", "anthropic/claude-sonnet-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -420,7 +420,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["zai-coding-plan"], model: "glm-5" },
         ],
         availableModels: new Set(["zai-coding-plan/glm-5", "opencode/glm-5"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -438,7 +438,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["zai-coding-plan"], model: "glm-5", variant: "high" },
         ],
         availableModels: new Set(["opencode/glm-5"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -457,7 +457,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "claude-sonnet-4-6" },
         ],
         availableModels: new Set(["anthropic/claude-sonnet-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -477,14 +477,14 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "nonexistent-model" },
         ],
         availableModels: new Set(["openai/gpt-5.2", "anthropic/claude-opus-4-6"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
       const result = resolveModelWithFallback(input)
 
       // then
-      expect(result!.model).toBe("google/gemini-3-pro")
+      expect(result!.model).toBe("google/gemini-3.1-pro")
       expect(result!.source).toBe("system-default")
       expect(logSpy).toHaveBeenCalledWith("No available model found in fallback chain, falling through to system default")
     })
@@ -516,7 +516,7 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic", "openai"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -533,7 +533,7 @@ describe("resolveModelWithFallback", () => {
       const cacheSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["github-copilot"])
       const input: ExtendedModelResolutionInput = {
         fallbackChain: [
-          { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro" },
+          { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro" },
         ],
         availableModels: new Set(),
         systemDefaultModel: "anthropic/claude-sonnet-4-6",
@@ -544,7 +544,7 @@ describe("resolveModelWithFallback", () => {
 
       // then - should use github-copilot (second provider) since google not connected
       // model name is transformed to preview variant for github-copilot provider
-      expect(result!.model).toBe("github-copilot/gemini-3-pro-preview")
+      expect(result!.model).toBe("github-copilot/gemini-3.1-pro-preview")
       expect(result!.source).toBe("provider-fallback")
       cacheSpy.mockRestore()
     })
@@ -577,14 +577,14 @@ describe("resolveModelWithFallback", () => {
           { providers: ["anthropic"], model: "claude-opus-4-6" },
         ],
         availableModels: new Set(),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
       const result = resolveModelWithFallback(input)
 
       // then - should fall through to system default
-      expect(result!.model).toBe("google/gemini-3-pro")
+      expect(result!.model).toBe("google/gemini-3.1-pro")
       expect(result!.source).toBe("system-default")
       cacheSpy.mockRestore()
     })
@@ -593,14 +593,14 @@ describe("resolveModelWithFallback", () => {
       // given
       const input: ExtendedModelResolutionInput = {
         availableModels: new Set(["openai/gpt-5.2"]),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
       const result = resolveModelWithFallback(input)
 
       // then
-      expect(result!.model).toBe("google/gemini-3-pro")
+      expect(result!.model).toBe("google/gemini-3.1-pro")
       expect(result!.source).toBe("system-default")
     })
   })
@@ -627,20 +627,20 @@ describe("resolveModelWithFallback", () => {
 
     test("tries all providers in first entry before moving to second entry", () => {
       // given
-      const availableModels = new Set(["google/gemini-3-pro"])
+      const availableModels = new Set(["google/gemini-3.1-pro"])
 
       // when
       const result = resolveModelWithFallback({
         fallbackChain: [
           { providers: ["openai", "anthropic"], model: "gpt-5.2" },
-          { providers: ["google"], model: "gemini-3-pro" },
+          { providers: ["google"], model: "gemini-3.1-pro" },
         ],
         availableModels,
         systemDefaultModel: "system/default",
       })
 
       // then
-      expect(result!.model).toBe("google/gemini-3-pro")
+      expect(result!.model).toBe("google/gemini-3.1-pro")
       expect(result!.source).toBe("provider-fallback")
     })
 
@@ -675,7 +675,7 @@ describe("resolveModelWithFallback", () => {
         fallbackChain: [
           { providers: ["openai"], model: "gpt-5.2" },
           { providers: ["anthropic"], model: "claude-opus-4-6" },
-          { providers: ["google"], model: "gemini-3-pro" },
+          { providers: ["google"], model: "gemini-3.1-pro" },
         ],
         availableModels,
         systemDefaultModel: "system/default",
@@ -693,7 +693,7 @@ describe("resolveModelWithFallback", () => {
       const input: ExtendedModelResolutionInput = {
         userModel: "anthropic/claude-opus-4-6",
         availableModels: new Set(),
-        systemDefaultModel: "google/gemini-3-pro",
+        systemDefaultModel: "google/gemini-3.1-pro",
       }
 
       // when
@@ -708,32 +708,32 @@ describe("resolveModelWithFallback", () => {
 
   describe("categoryDefaultModel (fuzzy matching for category defaults)", () => {
     test("applies fuzzy matching to categoryDefaultModel when userModel not provided", () => {
-      // given - gemini-3-pro is the category default, but only gemini-3-pro-preview is available
+      // given - gemini-3.1-pro is the category default, but only gemini-3.1-pro-preview is available
       const input: ExtendedModelResolutionInput = {
-        categoryDefaultModel: "google/gemini-3-pro",
+        categoryDefaultModel: "google/gemini-3.1-pro",
         fallbackChain: [
-          { providers: ["google", "github-copilot", "opencode"], model: "gemini-3-pro" },
+          { providers: ["google", "github-copilot", "opencode"], model: "gemini-3.1-pro" },
         ],
-        availableModels: new Set(["google/gemini-3-pro-preview", "anthropic/claude-opus-4-6"]),
+        availableModels: new Set(["google/gemini-3.1-pro-preview", "anthropic/claude-opus-4-6"]),
         systemDefaultModel: "anthropic/claude-sonnet-4-6",
       }
 
       // when
       const result = resolveModelWithFallback(input)
 
-      // then - should fuzzy match gemini-3-pro → gemini-3-pro-preview
-      expect(result!.model).toBe("google/gemini-3-pro-preview")
+      // then - should fuzzy match gemini-3.1-pro → gemini-3.1-pro-preview
+      expect(result!.model).toBe("google/gemini-3.1-pro-preview")
       expect(result!.source).toBe("category-default")
     })
 
     test("categoryDefaultModel uses exact match when available", () => {
       // given - exact match exists
       const input: ExtendedModelResolutionInput = {
-        categoryDefaultModel: "google/gemini-3-pro",
+        categoryDefaultModel: "google/gemini-3.1-pro",
         fallbackChain: [
-          { providers: ["google"], model: "gemini-3-pro" },
+          { providers: ["google"], model: "gemini-3.1-pro" },
         ],
-        availableModels: new Set(["google/gemini-3-pro", "google/gemini-3-pro-preview"]),
+        availableModels: new Set(["google/gemini-3.1-pro", "google/gemini-3.1-pro-preview"]),
         systemDefaultModel: "anthropic/claude-sonnet-4-6",
       }
 
@@ -741,14 +741,14 @@ describe("resolveModelWithFallback", () => {
       const result = resolveModelWithFallback(input)
 
       // then - should use exact match
-      expect(result!.model).toBe("google/gemini-3-pro")
+      expect(result!.model).toBe("google/gemini-3.1-pro")
       expect(result!.source).toBe("category-default")
     })
 
     test("categoryDefaultModel falls through to fallbackChain when no match in availableModels", () => {
       // given - categoryDefaultModel has no match, but fallbackChain does
       const input: ExtendedModelResolutionInput = {
-        categoryDefaultModel: "google/gemini-3-pro",
+        categoryDefaultModel: "google/gemini-3.1-pro",
         fallbackChain: [
           { providers: ["anthropic"], model: "claude-opus-4-6" },
         ],
@@ -768,11 +768,11 @@ describe("resolveModelWithFallback", () => {
       // given - both userModel and categoryDefaultModel provided
       const input: ExtendedModelResolutionInput = {
         userModel: "anthropic/claude-opus-4-6",
-        categoryDefaultModel: "google/gemini-3-pro",
+        categoryDefaultModel: "google/gemini-3.1-pro",
         fallbackChain: [
-          { providers: ["google"], model: "gemini-3-pro" },
+          { providers: ["google"], model: "gemini-3.1-pro" },
         ],
-        availableModels: new Set(["google/gemini-3-pro-preview", "anthropic/claude-opus-4-6"]),
+        availableModels: new Set(["google/gemini-3.1-pro-preview", "anthropic/claude-opus-4-6"]),
         systemDefaultModel: "system/default",
       }
 
@@ -788,7 +788,7 @@ describe("resolveModelWithFallback", () => {
       // given - no availableModels but connected provider cache exists
       const cacheSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["google"])
       const input: ExtendedModelResolutionInput = {
-        categoryDefaultModel: "google/gemini-3-pro",
+        categoryDefaultModel: "google/gemini-3.1-pro",
         availableModels: new Set(),
         systemDefaultModel: "anthropic/claude-sonnet-4-6",
       }
@@ -797,7 +797,7 @@ describe("resolveModelWithFallback", () => {
       const result = resolveModelWithFallback(input)
 
       // then - should use transformed categoryDefaultModel since google is connected
-      expect(result!.model).toBe("google/gemini-3-pro-preview")
+      expect(result!.model).toBe("google/gemini-3.1-pro-preview")
       expect(result!.source).toBe("category-default")
       cacheSpy.mockRestore()
     })
@@ -824,7 +824,7 @@ describe("resolveModelWithFallback", () => {
       // given - category default already has -preview suffix
       const cacheSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["google"])
       const input: ExtendedModelResolutionInput = {
-        categoryDefaultModel: "google/gemini-3-pro-preview",
+        categoryDefaultModel: "google/gemini-3.1-pro-preview",
         availableModels: new Set(),
         systemDefaultModel: "anthropic/claude-sonnet-4-5",
       }
@@ -832,18 +832,18 @@ describe("resolveModelWithFallback", () => {
       // when
       const result = resolveModelWithFallback(input)
 
-      // then - should NOT become gemini-3-pro-preview-preview
-      expect(result!.model).toBe("google/gemini-3-pro-preview")
+      // then - should NOT become gemini-3.1-pro-preview-preview
+      expect(result!.model).toBe("google/gemini-3.1-pro-preview")
       expect(result!.source).toBe("category-default")
       cacheSpy.mockRestore()
     })
 
-    test("transforms gemini-3-pro in fallback chain for google connected provider", () => {
-      // given - google connected, fallback chain has gemini-3-pro
+    test("transforms gemini-3.1-pro in fallback chain for google connected provider", () => {
+      // given - google connected, fallback chain has gemini-3.1-pro
       const cacheSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["google"])
       const input: ExtendedModelResolutionInput = {
         fallbackChain: [
-          { providers: ["google", "github-copilot"], model: "gemini-3-pro" },
+          { providers: ["google", "github-copilot"], model: "gemini-3.1-pro" },
         ],
         availableModels: new Set(),
         systemDefaultModel: "anthropic/claude-sonnet-4-5",
@@ -853,7 +853,7 @@ describe("resolveModelWithFallback", () => {
       const result = resolveModelWithFallback(input)
 
       // then - should transform to preview variant for google provider
-      expect(result!.model).toBe("google/gemini-3-pro-preview")
+      expect(result!.model).toBe("google/gemini-3.1-pro-preview")
       expect(result!.source).toBe("provider-fallback")
       cacheSpy.mockRestore()
     })

src/shared/provider-model-id-transform.ts

@@ -6,12 +6,12 @@ export function transformModelForProvider(provider: string, model: string): stri
 			.replace("claude-sonnet-4-5", "claude-sonnet-4.5")
 			.replace("claude-haiku-4-5", "claude-haiku-4.5")
 			.replace("claude-sonnet-4", "claude-sonnet-4")
-			.replace(/gemini-3-pro(?!-)/g, "gemini-3-pro-preview")
+			.replace(/gemini-3\.1-pro(?!-)/g, "gemini-3.1-pro-preview")
 			.replace(/gemini-3-flash(?!-)/g, "gemini-3-flash-preview")
 	}
 	if (provider === "google") {
 		return model
-			.replace(/gemini-3-pro(?!-)/g, "gemini-3-pro-preview")
+			.replace(/gemini-3\.1-pro(?!-)/g, "gemini-3.1-pro-preview")
 			.replace(/gemini-3-flash(?!-)/g, "gemini-3-flash-preview")
 	}
 	return model

src/shared/spawn-with-windows-hide.ts

@@ -0,0 +1,84 @@
+import { spawn as bunSpawn } from "bun"
+import { spawn as nodeSpawn, type ChildProcess } from "node:child_process"
+import { Readable } from "node:stream"
+
+export interface SpawnOptions {
+  cwd?: string
+  env?: Record<string, string | undefined>
+  stdin?: "pipe" | "inherit" | "ignore"
+  stdout?: "pipe" | "inherit" | "ignore"
+  stderr?: "pipe" | "inherit" | "ignore"
+}
+
+export interface SpawnedProcess {
+  readonly exitCode: number | null
+  readonly exited: Promise<number>
+  readonly stdout: ReadableStream<Uint8Array> | undefined
+  readonly stderr: ReadableStream<Uint8Array> | undefined
+  kill(signal?: NodeJS.Signals): void
+}
+
+function toReadableStream(stream: NodeJS.ReadableStream | null): ReadableStream<Uint8Array> | undefined {
+  if (!stream) {
+    return undefined
+  }
+
+  return Readable.toWeb(stream as Readable) as ReadableStream<Uint8Array>
+}
+
+function wrapNodeProcess(proc: ChildProcess): SpawnedProcess {
+  let resolveExited: (exitCode: number) => void
+  let exitCode: number | null = null
+
+  const exited = new Promise<number>((resolve) => {
+    resolveExited = resolve
+  })
+
+  proc.on("exit", (code) => {
+    exitCode = code ?? 1
+    resolveExited(exitCode)
+  })
+
+  proc.on("error", () => {
+    if (exitCode === null) {
+      exitCode = 1
+      resolveExited(1)
+    }
+  })
+
+  return {
+    get exitCode() {
+      return exitCode
+    },
+    exited,
+    stdout: toReadableStream(proc.stdout),
+    stderr: toReadableStream(proc.stderr),
+    kill(signal?: NodeJS.Signals): void {
+      try {
+        if (!signal) {
+          proc.kill()
+          return
+        }
+
+        proc.kill(signal)
+      } catch {}
+    },
+  }
+}
+
+export function spawnWithWindowsHide(command: string[], options: SpawnOptions): SpawnedProcess {
+  if (process.platform !== "win32") {
+    return bunSpawn(command, options)
+  }
+
+  const [cmd, ...args] = command
+  const proc = nodeSpawn(cmd, args, {
+    cwd: options.cwd,
+    env: options.env,
+    stdio: [options.stdin ?? "pipe", options.stdout ?? "pipe", options.stderr ?? "pipe"],
+    windowsHide: true,
+    shell: true,
+  })
+
+  return wrapNodeProcess(proc)
+}

src/tools/delegate-task/constants.ts

@@ -208,10 +208,10 @@ You are NOT an interactive assistant. You are an autonomous problem-solver.
 
 
 export const DEFAULT_CATEGORIES: Record<string, CategoryConfig> = {
-  "visual-engineering": { model: "google/gemini-3-pro", variant: "high" },
+  "visual-engineering": { model: "google/gemini-3.1-pro", variant: "high" },
   ultrabrain: { model: "openai/gpt-5.3-codex", variant: "xhigh" },
   deep: { model: "openai/gpt-5.3-codex", variant: "medium" },
-  artistry: { model: "google/gemini-3-pro", variant: "high" },
+  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" },

src/tools/delegate-task/token-limiter.test.ts

@@ -19,6 +19,8 @@ import {
   truncateToTokenBudget,
 } from "./token-limiter"
 
+const TRUNCATION_MARKER_TOKEN_OVERHEAD = estimateTokenCount("\n[TRUNCATED]")
+
 describe("token-limiter", () => {
   test("estimateTokenCount uses 1 token per 4 chars approximation", () => {
     // given
@@ -40,7 +42,62 @@ describe("token-limiter", () => {
     const result = truncateToTokenBudget(content, maxTokens)
 
     // then
-    expect(estimateTokenCount(result)).toBeLessThanOrEqual(maxTokens)
+    expect(estimateTokenCount(result)).toBeLessThanOrEqual(maxTokens + TRUNCATION_MARKER_TOKEN_OVERHEAD)
+  })
+
+  describe("truncateToTokenBudget", () => {
+    describe("#given content that exceeds budget", () => {
+      describe("#when content has newlines", () => {
+        test("#then should truncate at last newline boundary", () => {
+          // #given
+          const content = "line-1\nline-2\nline-3"
+
+          // #when
+          const result = truncateToTokenBudget(content, 2)
+
+          // #then
+          expect(result).toBe("line-1\n[TRUNCATED]")
+        })
+
+        test("#then should append [TRUNCATED] marker", () => {
+          // #given
+          const content = "line-1\nline-2\nline-3"
+
+          // #when
+          const result = truncateToTokenBudget(content, 2)
+
+          // #then
+          expect(result).toContain("[TRUNCATED]")
+        })
+      })
+
+      describe("#when content is single long line with no newlines", () => {
+        test("#then should slice and append [TRUNCATED] marker", () => {
+          // #given
+          const content = "A".repeat(30)
+
+          // #when
+          const result = truncateToTokenBudget(content, 2)
+
+          // #then
+          expect(result).toBe("AAAAAAAA\n[TRUNCATED]")
+        })
+      })
+    })
+
+    describe("#given content within budget", () => {
+      test("#then should return content unchanged without marker", () => {
+        // #given
+        const content = "line-1\nline-2"
+
+        // #when
+        const result = truncateToTokenBudget(content, 20)
+
+        // #then
+        expect(result).toBe(content)
+        expect(result).not.toContain("[TRUNCATED]")
+      })
+    })
   })
 
   test("buildSystemContentWithTokenLimit returns undefined when there is no content", () => {
@@ -76,10 +133,10 @@ describe("token-limiter", () => {
     const result = buildSystemContentWithTokenLimit(input, 80)
 
     // then
-    expect(result).toContain("AGENTS_CONTEXT:keep")
-    expect(result).toContain("CATEGORY_APPEND:keep")
+    expect(result).toContain("AGENTS_C")
+    expect(result).toContain("CATE")
     expect(result).toContain("SKILL_ALPHA:")
-    expect(estimateTokenCount(result as string)).toBeLessThanOrEqual(80)
+    expect(estimateTokenCount(result as string)).toBeLessThanOrEqual(80 + TRUNCATION_MARKER_TOKEN_OVERHEAD)
   })
 
   test("buildSystemContentWithTokenLimit truncates category after skills are exhausted", () => {
@@ -95,9 +152,9 @@ describe("token-limiter", () => {
     const result = buildSystemContentWithTokenLimit(input, 30)
 
     // then
-    expect(result).toContain("AGENTS_CONTEXT:keep")
+    expect(result).toContain("AGENTS_C")
     expect(result).not.toContain("SKILL_ALPHA:" + "a".repeat(80))
-    expect(estimateTokenCount(result as string)).toBeLessThanOrEqual(30)
+    expect(estimateTokenCount(result as string)).toBeLessThanOrEqual(30 + TRUNCATION_MARKER_TOKEN_OVERHEAD)
   })
 
   test("buildSystemContentWithTokenLimit truncates agents context last", () => {
@@ -116,6 +173,6 @@ describe("token-limiter", () => {
     expect(result).toContain("AGENTS_CONTEXT:")
     expect(result).not.toContain("SKILL_ALPHA:")
     expect(result).not.toContain("CATEGORY_APPEND:")
-    expect(estimateTokenCount(result as string)).toBeLessThanOrEqual(10)
+    expect(estimateTokenCount(result as string)).toBeLessThanOrEqual(10 + TRUNCATION_MARKER_TOKEN_OVERHEAD)
   })
 })

src/tools/delegate-task/token-limiter.ts

@@ -20,7 +20,13 @@ export function truncateToTokenBudget(content: string, maxTokens: number): strin
     return content
   }
 
-  return content.slice(0, maxCharacters)
+  const sliced = content.slice(0, maxCharacters)
+  const lastNewline = sliced.lastIndexOf("\n")
+  if (lastNewline > 0) {
+    return `${sliced.slice(0, lastNewline)}\n[TRUNCATED]`
+  }
+
+  return `${sliced}\n[TRUNCATED]`
 }
 
 function joinSystemParts(parts: string[]): string | undefined {

src/tools/delegate-task/tools.test.ts

@@ -17,7 +17,7 @@ const TEST_AVAILABLE_MODELS = new Set([
   "anthropic/claude-opus-4-6",
   "anthropic/claude-sonnet-4-6",
   "anthropic/claude-haiku-4-5",
-  "google/gemini-3-pro",
+  "google/gemini-3.1-pro",
   "google/gemini-3-flash",
   "openai/gpt-5.2",
   "openai/gpt-5.3-codex",
@@ -52,7 +52,7 @@ describe("sisyphus-task", () => {
     providerModelsSpy = spyOn(connectedProvidersCache, "readProviderModelsCache").mockReturnValue({
       models: {
         anthropic: ["claude-opus-4-6", "claude-sonnet-4-6", "claude-haiku-4-5"],
-        google: ["gemini-3-pro", "gemini-3-flash"],
+        google: ["gemini-3.1-pro", "gemini-3-flash"],
         openai: ["gpt-5.2", "gpt-5.3-codex"],
       },
       connected: ["anthropic", "google", "openai"],
@@ -73,7 +73,7 @@ describe("sisyphus-task", () => {
 
       // when / #then
       expect(category).toBeDefined()
-      expect(category.model).toBe("google/gemini-3-pro")
+      expect(category.model).toBe("google/gemini-3.1-pro")
       expect(category.variant).toBe("high")
     })
 
@@ -781,7 +781,7 @@ describe("sisyphus-task", () => {
 
       // then
       expect(result).not.toBeNull()
-      expect(result!.config.model).toBe("google/gemini-3-pro")
+      expect(result!.config.model).toBe("google/gemini-3.1-pro")
       expect(result!.promptAppend).toContain("VISUAL/UI")
     })
 
@@ -805,7 +805,7 @@ describe("sisyphus-task", () => {
       const categoryName = "visual-engineering"
       const userCategories = {
         "visual-engineering": {
-          model: "google/gemini-3-pro",
+          model: "google/gemini-3.1-pro",
           prompt_append: "Custom instructions here",
         },
       }
@@ -845,7 +845,7 @@ describe("sisyphus-task", () => {
       const categoryName = "visual-engineering"
       const userCategories = {
         "visual-engineering": {
-          model: "google/gemini-3-pro",
+          model: "google/gemini-3.1-pro",
           temperature: 0.3,
         },
       }
@@ -868,7 +868,7 @@ describe("sisyphus-task", () => {
 
       // then - category's built-in model wins over inheritedModel
       expect(result).not.toBeNull()
-      expect(result!.config.model).toBe("google/gemini-3-pro")
+      expect(result!.config.model).toBe("google/gemini-3.1-pro")
     })
 
     test("systemDefaultModel is used as fallback when custom category has no model", () => {
@@ -910,7 +910,7 @@ describe("sisyphus-task", () => {
 
       // then
       expect(result).not.toBeNull()
-      expect(result!.config.model).toBe("google/gemini-3-pro")
+      expect(result!.config.model).toBe("google/gemini-3.1-pro")
     })
   })
 
@@ -1738,7 +1738,7 @@ describe("sisyphus-task", () => {
        const mockClient = {
          app: { agents: async () => ({ data: [] }) },
          config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
-         model: { list: async () => [{ provider: "google", id: "gemini-3-pro" }] },
+         model: { list: async () => [{ provider: "google", id: "gemini-3.1-pro" }] },
          session: {
            get: async () => ({ data: { directory: "/project" } }),
            create: async () => ({ data: { id: "ses_unstable_gemini" } }),
@@ -2001,7 +2001,7 @@ describe("sisyphus-task", () => {
        const mockClient = {
          app: { agents: async () => ({ data: [] }) },
          config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
-         model: { list: async () => [{ provider: "google", id: "gemini-3-pro" }] },
+         model: { list: async () => [{ provider: "google", id: "gemini-3.1-pro" }] },
          session: {
            get: async () => ({ data: { directory: "/project" } }),
            create: async () => ({ data: { id: "ses_artistry_gemini" } }),
@@ -2028,7 +2028,7 @@ describe("sisyphus-task", () => {
         abort: new AbortController().signal,
       }
       
-      // when - artistry category (gemini-3-pro with high variant)
+      // when - artistry category (gemini-3.1-pro with high variant)
       const result = await tool.execute(
         {
           description: "Test artistry forced background",
@@ -3026,9 +3026,9 @@ describe("sisyphus-task", () => {
       // when resolveCategoryConfig is called
       const resolved = resolveCategoryConfig(categoryName, { userCategories, inheritedModel, systemDefaultModel: SYSTEM_DEFAULT_MODEL })
       
-      // then should use category's built-in model (gemini-3-pro for visual-engineering)
+      // then should use category's built-in model (gemini-3.1-pro for visual-engineering)
       expect(resolved).not.toBeNull()
-      expect(resolved!.model).toBe("google/gemini-3-pro")
+      expect(resolved!.model).toBe("google/gemini-3.1-pro")
     })
 
     test("systemDefaultModel is used when no other model is available", () => {
@@ -3522,7 +3522,7 @@ describe("sisyphus-task", () => {
       )
 
       // then - should resolve via AGENT_MODEL_REQUIREMENTS fallback chain for oracle
-      // oracle fallback chain: gpt-5.2 (openai) > gemini-3-pro (google) > claude-opus-4-6 (anthropic)
+      // oracle fallback chain: gpt-5.2 (openai) > gemini-3.1-pro (google) > claude-opus-4-6 (anthropic)
       // Since openai is in connectedProviders, should resolve to openai/gpt-5.2
       expect(promptBody.model).toBeDefined()
       expect(promptBody.model.providerID).toBe("openai")

src/tools/hashline-edit/diff-utils.test.ts

@@ -1,5 +1,6 @@
 /// <reference types="bun-types" />
 import { describe, expect, it } from "bun:test"
+import { parsePatch } from "diff"
 import { generateUnifiedDiff } from "./diff-utils"
 
 function createNumberedLines(totalLineCount: number): string {
@@ -7,6 +8,66 @@ function createNumberedLines(totalLineCount: number): string {
 }
 
 describe("generateUnifiedDiff", () => {
+  describe("#given OpenCode compatibility format", () => {
+    it("#then includes the Index header emitted by diff library", () => {
+      //#given
+      const oldContent = "a\n"
+      const newContent = "b\n"
+
+      //#when
+      const diff = generateUnifiedDiff(oldContent, newContent, "test.ts")
+
+      //#then
+      expect(diff).toContain("Index: test.ts")
+    })
+
+    it("#then includes unified --- and +++ file headers", () => {
+      //#given
+      const oldContent = "a\n"
+      const newContent = "b\n"
+
+      //#when
+      const diff = generateUnifiedDiff(oldContent, newContent, "test.ts")
+
+      //#then
+      expect(diff).toContain("--- test.ts")
+      expect(diff).toContain("+++ test.ts")
+    })
+
+    it("#then remains parseable by OpenCode parsePatch flow", () => {
+      //#given
+      const oldContent = "line1\nline2\n"
+      const newContent = "line1\nline2-updated\n"
+
+      //#when
+      const diff = generateUnifiedDiff(oldContent, newContent, "test.ts")
+      const patches = parsePatch(diff)
+
+      //#then
+      expect(patches).toHaveLength(1)
+      expect(patches[0]?.oldFileName).toBe("test.ts")
+      expect(patches[0]?.newFileName).toBe("test.ts")
+      expect(patches[0]?.hunks).toHaveLength(1)
+    })
+  })
+
+  describe("#given content without trailing newline", () => {
+    it("#then keeps no-newline markers parseable", () => {
+      //#given
+      const oldContent = "a"
+      const newContent = "b"
+
+      //#when
+      const diff = generateUnifiedDiff(oldContent, newContent, "test.ts")
+      const patches = parsePatch(diff)
+      const hunkLines = patches[0]?.hunks[0]?.lines ?? []
+
+      //#then
+      expect(diff).toContain("\\ No newline at end of file")
+      expect(hunkLines).toEqual(["-a", "\\ No newline at end of file", "+b", "\\ No newline at end of file"])
+    })
+  })
+
   it("creates separate hunks for distant changes", () => {
     //#given
     const oldContent = createNumberedLines(60)

src/tools/hashline-edit/edit-deduplication.ts

@@ -1,18 +1,24 @@
 import type { HashlineEdit } from "./types"
 import { toNewLines } from "./edit-text-normalization"
+import { normalizeLineRef } from "./validation"
 
 function normalizeEditPayload(payload: string | string[]): string {
   return toNewLines(payload).join("\n")
 }
 
+function canonicalAnchor(anchor: string | undefined): string {
+  if (!anchor) return ""
+  return normalizeLineRef(anchor)
+}
+
 function buildDedupeKey(edit: HashlineEdit): string {
   switch (edit.op) {
     case "replace":
-      return `replace|${edit.pos}|${edit.end ?? ""}|${normalizeEditPayload(edit.lines)}`
+      return `replace|${canonicalAnchor(edit.pos)}|${edit.end ? canonicalAnchor(edit.end) : ""}|${normalizeEditPayload(edit.lines)}`
     case "append":
-      return `append|${edit.pos ?? ""}|${normalizeEditPayload(edit.lines)}`
+      return `append|${canonicalAnchor(edit.pos)}|${normalizeEditPayload(edit.lines)}`
     case "prepend":
-      return `prepend|${edit.pos ?? ""}|${normalizeEditPayload(edit.lines)}`
+      return `prepend|${canonicalAnchor(edit.pos)}|${normalizeEditPayload(edit.lines)}`
     default:
       return JSON.stringify(edit)
   }

src/tools/hashline-edit/edit-operations.test.ts

@@ -1,5 +1,5 @@
 import { describe, expect, it } from "bun:test"
-import { applyHashlineEdits } from "./edit-operations"
+import { applyHashlineEdits, applyHashlineEditsWithReport } from "./edit-operations"
 import { applyAppend, applyInsertAfter, applyPrepend, applyReplaceLines, applySetLine } from "./edit-operation-primitives"
 import { computeLineHash } from "./hash-computation"
 import type { HashlineEdit } from "./types"
@@ -389,3 +389,23 @@ describe("hashline edit operations", () => {
     expect(result).toEqual("replaced A\nline 3\nreplaced B")
   })
 })
+
+describe("dedupe anchor canonicalization", () => {
+  it("deduplicates edits with whitespace-variant anchors", () => {
+    //#given
+    const content = "line 1\nline 2"
+    const lines = content.split("\n")
+    const canonical = `1#${computeLineHash(1, lines[0])}`
+    const spaced = ` 1 # ${computeLineHash(1, lines[0])} `
+
+    //#when
+    const report = applyHashlineEditsWithReport(content, [
+      { op: "append", pos: canonical, lines: ["inserted"] },
+      { op: "append", pos: spaced, lines: ["inserted"] },
+    ])
+
+    //#then
+    expect(report.deduplicatedEdits).toBe(1)
+    expect(report.content).toBe("line 1\ninserted\nline 2")
+  })
+})

src/tools/hashline-edit/hashline-edit-executor.ts

@@ -33,7 +33,7 @@ function resolveToolCallID(ctx: ToolContextWithCallID): string | undefined {
 
 function canCreateFromMissingFile(edits: HashlineEdit[]): boolean {
   if (edits.length === 0) return false
-  return edits.every((edit) => edit.op === "append" || edit.op === "prepend")
+  return edits.every((edit) => (edit.op === "append" || edit.op === "prepend") && !edit.pos)
 }
 
 function buildSuccessMeta(
@@ -86,19 +86,19 @@ export async function executeHashlineEditTool(args: HashlineEditArgs, context: T
     const filePath = args.filePath
     const { delete: deleteMode, rename } = args
 
+    if (deleteMode && rename) {
+      return "Error: delete and rename cannot be used together"
+    }
+    if (deleteMode && args.edits.length > 0) {
+      return "Error: delete mode requires edits to be an empty array"
+    }
+
     if (!deleteMode && (!args.edits || !Array.isArray(args.edits) || args.edits.length === 0)) {
       return "Error: edits parameter must be a non-empty array"
     }
 
     const edits = deleteMode ? [] : normalizeHashlineEdits(args.edits)
 
-    if (deleteMode && rename) {
-      return "Error: delete and rename cannot be used together"
-    }
-    if (deleteMode && edits.length > 0) {
-      return "Error: delete mode requires edits to be an empty array"
-    }
-
     const file = Bun.file(filePath)
     const exists = await file.exists()
     if (!exists && !deleteMode && !canCreateFromMissingFile(edits)) {

src/tools/hashline-edit/tool-description.ts

@@ -10,7 +10,7 @@ WORKFLOW:
 VALIDATION:
   Payload shape: { "filePath": string, "edits": [...], "delete"?: boolean, "rename"?: string }
   Each edit must be one of: replace, append, prepend
-  Edit shape: { "op": "replace"|"append"|"prepend", "pos"?: "LINE#ID", "end"?: "LINE#ID", "lines"?: string|string[]|null }
+  Edit shape: { "op": "replace"|"append"|"prepend", "pos"?: "LINE#ID", "end"?: "LINE#ID", "lines": string|string[]|null }
   lines must contain plain replacement text only (no LINE#ID prefixes, no diff + markers)
   CRITICAL: all operations validate against the same pre-edit file snapshot and apply bottom-up. Refs/tags are interpreted against the last-read version of the file.
 

src/tools/hashline-edit/tools.test.ts

@@ -341,4 +341,81 @@ describe("createHashlineEditTool", () => {
     //#then
     expect(envelope.lineEnding).toBe("\r\n")
   })
+
+  it("rejects delete=true with non-empty edits before normalization", async () => {
+    //#given
+    const filePath = path.join(tempDir, "delete-reject.txt")
+    fs.writeFileSync(filePath, "line1")
+
+    //#when
+    const result = await tool.execute(
+      {
+        filePath,
+        delete: true,
+        edits: [{ op: "replace", pos: "1#ZZ", lines: "bad" }],
+      },
+      createMockContext(),
+    )
+
+    //#then
+    expect(result).toContain("delete mode requires edits to be an empty array")
+    expect(fs.existsSync(filePath)).toBe(true)
+  })
+
+  it("rejects delete=true combined with rename", async () => {
+    //#given
+    const filePath = path.join(tempDir, "delete-rename.txt")
+    fs.writeFileSync(filePath, "line1")
+
+    //#when
+    const result = await tool.execute(
+      {
+        filePath,
+        delete: true,
+        rename: path.join(tempDir, "new-name.txt"),
+        edits: [],
+      },
+      createMockContext(),
+    )
+
+    //#then
+    expect(result).toContain("delete and rename cannot be used together")
+    expect(fs.existsSync(filePath)).toBe(true)
+  })
+
+  it("rejects missing file creation with anchored append", async () => {
+    //#given
+    const filePath = path.join(tempDir, "nonexistent.txt")
+
+    //#when
+    const result = await tool.execute(
+      {
+        filePath,
+        edits: [{ op: "append", pos: "1#ZZ", lines: ["bad"] }],
+      },
+      createMockContext(),
+    )
+
+    //#then
+    expect(result).toContain("File not found")
+  })
+
+  it("allows missing file creation with unanchored append", async () => {
+    //#given
+    const filePath = path.join(tempDir, "newfile.txt")
+
+    //#when
+    const result = await tool.execute(
+      {
+        filePath,
+        edits: [{ op: "append", lines: ["created"] }],
+      },
+      createMockContext(),
+    )
+
+    //#then
+    expect(fs.existsSync(filePath)).toBe(true)
+    expect(fs.readFileSync(filePath, "utf-8")).toBe("created")
+    expect(result).toBe(`Updated ${filePath}`)
+  })
 })

src/tools/hashline-edit/tools.ts

@@ -31,7 +31,6 @@ export function createHashlineEditTool(): ToolDefinition {
             end: tool.schema.string().optional().describe("Range end anchor in LINE#ID format"),
             lines: tool.schema
               .union([tool.schema.string(), tool.schema.array(tool.schema.string()), tool.schema.null()])
-              .optional()
               .describe("Replacement or inserted lines. null/[] deletes with replace"),
           })
         )

src/tools/hashline-edit/validation.ts

@@ -15,7 +15,7 @@ const MISMATCH_CONTEXT = 2
 
 const LINE_REF_EXTRACT_PATTERN = /([0-9]+#[ZPMQVRWSNKTXJBYH]{2})/
 
-function normalizeLineRef(ref: string): string {
+export function normalizeLineRef(ref: string): string {
   const originalTrimmed = ref.trim()
   let trimmed = originalTrimmed
   trimmed = trimmed.replace(/^(?:>>>|[+-])\s*/, "")