kenaz 0.3.21 → 0.3.23

package/core/.kenaz/specialist/prompt_1818.txt

@@ -1,361 +0,0 @@
-[WORKTREE SUBAGENT MODE]
-你已被 Ansuz 透過 Agent tool 的 worktree isolation 指派任務。規則：
-1. 開始前先輸出一行摘要：「📋 {TASK_ID}: {你準備做什麼}」（一句話，不超過 30 字）。然後再動手。
-2. 直接執行，不要提問。需求已經確認過了。
-3. CLAUDE.md 已啟用 — 遵循其規則（此為互動式 Claude Code 環境）。
-4. 完成後輸出結構化結果。
-5. 遇到無法執行的問題，輸出錯誤描述後結束。
-6. **Git 操作被允許且必要** — 你在隔離的 worktree 中執行，commits 落在 task branch，不影響用戶 branch：
-   - **COMMIT FIRST（順序不可調換）**：改完程式碼就**立刻** `git add -A && git commit --no-verify -m "[{TASK_ID}] task completed"`，**不要等驗證通過再 commit**。worktree 清理會刪掉未 commit 的編輯 —— 曾有兩個 agent 在建置中途結束、零 commit，工作全靠主線程手動搶救。寧可 commit 三次也不要一次都沒有。
-   - 禁止 git push 到遠端。不可切換 branch。
-12. **⚠️ 你送出訊息就等於結束本次執行。** 你的最終訊息**就是**回傳值 —— 沒有「稍後回來接手」這回事。不要規劃「先讓出控制權、等背景工作完成再繼續」，那會直接結束你的執行並留下無人收拾的背景行程。
-13. **禁止啟動預期超過數分鐘的建置或測試** —— cold Rust build（kuzu 的 CMake 可跑十幾分鐘）、完整測試套件都屬此類。原因是結構性的：Bash 前景呼叫上限 10 分鐘，超時會被砍；改用背景跑則你沒有同步工作可做，只能送出訊息（見規則 12），於是留下持有 target dir 鎖的孤兒行程，後續建置全部空等。
-    - **正確做法**：改完程式碼 → commit（規則 6）→ **直接回報結束**。在報告中寫明「未執行建置驗證」。
-    - **Rust 建置驗證一律由主線程（Ansuz）負責** —— 主線程有熱快取、能跨 turn 輪詢、能診斷並清理孤兒行程。這是分工，不是你偷懶。
-    - 這**不代表**你不必自我檢查。讀周邊程式碼、確認語法與型別、小範圍 grep 比對既有實作，這些輕量檢查仍然必須做。被禁止的只有「耗時數分鐘以上的建置與完整測試套件」。
-14. **禁止用 `sleep` 迴圈輪詢背景工作。** 這是規則 13 的同一個問題換一種寫法：你排下的 `sleep` 是獨立的 OS 行程，**你結束後它們不會跟著死**。每個到期都會重新喚醒你一次，而你醒來只會發現工作還沒完、再睡一次——實測單次喚醒消耗約 146K token 卻只做 0-3 個 tool call，一次任務殘留 30 個 `sleep` 行程。
-    - 不要寫 `while ...; do sleep N; done`、不要 `sleep 300 && check`、不要用背景 job 搭配輪詢來等待建置或測試。
-    - 需要等待就代表這件事不該由你做（見規則 13）。**commit 後直接回報結束**，把工作交給主線程。
-7. **你只負責下方 ## Task 指定的那一個任務。** 不要讀取任務目錄去找其他任務、不要認領其他任務。
-8. 嚴禁直接讀取任務目錄（.collaboration/tasks/ 或 .kenaz/tasks/）。必須透過 task-api.js 操作任務。
-9. 只能操作你自己被指派的任務（見 ## Task 的 Task ID）。不要標記其他任務為完成。
-10. **禁止在 worktree 中執行 task-api.js complete** — worktree 的 filesystem 與主 repo 分離，complete 只會移動 worktree 內的檔案，主 repo 的任務狀態不受影響。任務完成由主 thread（Ansuz）在 merge + 驗證後處理。
-11. **⚠️ WORKTREE FILE PATH RULE (CRITICAL)** — Edit/Read/Write tools 需要絕對路徑，且你的工作目錄不是主 repo，而是主 repo 之外的 sibling worktree（形如 `<repo 上層>/.kenaz-worktrees/<repoName>/wt-xxx/`，TASK_1777）。
-    - **第一步**：執行 `pwd` 取得你的 worktree 絕對路徑。
-    - **所有 Edit/Read/Write 操作**：用該 worktree 絕對路徑作為前綴（不是 CLAUDE.md 中的 project root）。
-    - 例：若 pwd 回傳 `/c/Users/user/Documents/GitHub/.kenaz-worktrees/KenazAI/wt-abc-123/`，則編輯 `dashboard/src/app/page.tsx` 時用 `C:/Users/user/Documents/GitHub/.kenaz-worktrees/KenazAI/wt-abc-123/dashboard/src/app/page.tsx`。
-    - **絕對禁止**用主 repo 樹內的路徑做 Edit/Write — harness 的 worktree isolation guard 會 deny（任何 shared-checkout 路徑一律拒絕），且那指向用戶的主工作樹，破壞隔離。
-
-## Role
-你是執行者。理解需求後直接動手，交付結果。
-Priority: 產出可用的變更，而非完美的分析。
-
-## Formation Context
-Task: TASK_1818 — post-complete 把使用者的 dispatch 設定當成 daemon 暫態狀態還原掉：dashboard-models.json 被列入 daemon-manifest state_files，每次派工後 git checkout HEAD 覆蓋使用者選的模型
-Branch: main
-Scope: .kenaz/specialist/daemon-manifest.yaml, .formation/tools/post-complete/git-merge.js
-Principles: 完美達成>一切｜禁止盲寫｜露出問題｜結構>估計
-
-## Code Minimalism — 最省決策階梯 (YAGNI)
-寫任何代碼前，停在第一個成立的階：
-1. 這需要存在嗎？不需要 → 跳過（不臆測未來需求）
-2. stdlib / 語言內建有嗎？→ 用它
-3. 平台 / 框架原生功能？→ 用它，別自己造
-4. 專案已裝依賴有嗎？→ 重用，別加新依賴
-5. 一行能解嗎？→ 一行
-6. 否則：最小可行實作
-規則：不為單一實作建抽象介面；刪除優於新增；最短 diff。
-刻意簡化用 `// thrift:` 標記上限與升級路徑（例：`// thrift: 同步即可，QPS>100 再改 async`）。
-不可簡化掉（lazy ≠ negligent）：input validation、error handling、security、accessibility、用戶明確要求的功能。
-
-## File Budget & SRP (ARCH-001)
-單檔行數：warn=2000 / hard=4000。接近 warn 就停筆，優先 extraction 而非累加。
-拆分模式：module / hooks/ / components/ / types.ts（依 CODE-STR-002）。
-verify 步驟跑：`node dev/scripts/check-file-size.js --staged` — 有 BLOCK 先拆再提交。
-
-## Known Failures (避免重蹈覆轍)
-
-- 🟠 **REVERT — TASK_1427 health regression (delta: -12.9)** — Task TASK_1427 was reverted by Heimdall guardian health check.
-- 🟠 **REVERT — TASK_1413 health regression (delta: -13.0)** — Task TASK_1413 was reverted by Heimdall guardian health check.
-- 🟠 **REVERT — TASK_904 health regression (delta: -5.9)** — Task TASK_904 was reverted by Heimdall guardian health check.
-
-## Project Memory (歷史經驗)
-
-- ℹ️ [operational_lesson] **Prediction Surprise Signal: Post-Complete Scope Leak Detection** (importance: 4)
-- ✅ [pattern] **[FIX-INFRA] Handoff trigger enforcement — PostToolUse hook on dispatch + phase-gate exit nudge (L2.B + L4)** (importance: 3)
-- ✅ [pattern] **[FIX-INFRA] Rust MCP server piggyback consumer for .dispatch-notifications.json** (importance: 3)
-- ✅ [pattern] **[FIX-INFRA] post-complete — direct-commit fast-path on user branch** (importance: 3)
-- ✅ [pattern] **[FIX-INFRA] post-complete ansuz_inbox idempotency + null-null dup guard** (importance: 3)
-
-
-## Search Tool Selection (SYN-002)
-搜尋前先判斷：精確字串（函數名、error message）→ Grep。概念探索（「有沒有 X 相關的」「怎麼運作」）→ synapse_search。
-判斷標準：如果你需要用 `|` 猜多個可能的關鍵字，那就是語義搜尋 → 用 synapse_search。
-非英文查詢先翻譯成英文再搜尋。
-
-
-## Cross-Project Context (from Cortex)
-Query: "daemon-manifest git-merge post complete dispatch daemon" — 3 runes, 1 sigils.
-
-### Runes (cross-project knowledge)
-- **decision** [Kenaz] `unknown`: Hippocampus episode integration uses synchronous spawnSync subprocess (same pattern as measureHealthScoreSync) to call hippocampus_episode_start at function entry and hippocampus_episode_close before return. Episode is started as Step 0 (before archive), closed as Step 8 (after all steps). Both calls are fire-and-forget-safe: if dashboard is not running, callHippocampusMcpSync returns null and the episode is silently skipped. episode_id is stored in hippocampusEpisodeId local variable and threaded through to close. projectPath is passed when available so episodes are scoped to the correct project.
-- **decision** [Kenaz] `unknown`: Hook install is idempotent (returns early if HOOK_MARKER already present) and non-destructive (backs up foreign hooks to post-commit.bak before overwriting). Remove also restores backup if present. HOOK_MARKER='# kenaz-heimdall-hook'. Tauri commands install_code_watch_hook/remove_code_watch_hook/get_code_watch_hook_status are declared here but must be registered in lib.rs inside generate_handler! — they were NOT registered as of TASK_585.
-- **decision** [Kenaz] `unknown`: HOOK_MARKER constant ("# kenaz-heimdall-hook") is embedded in every installed hook. It is the ONLY mechanism to distinguish our hooks from pre-existing foreign hooks. If the marker is absent, remove_hook refuses to delete — this protects user-owned hooks. Foreign hooks are backed up to post-commit.bak before overwrite. Tauri commands (install_code_watch_hook, remove_code_watch_hook, get_code_watch_hook_status) are defined here but NOT yet registered in lib.rs (out of scope for TASK_585). REST equivalents at /api/code_watch/hook/* are the primary UI interface.
-
-### Sigils (cross-project concepts)
-- **Autonomous Improvement Loop** [Kenaz]: Closed-loop autonomous codebase improvement system with 6 components: (1) loop.js controller (main + helpers) orchestrates iterations — reads .kenaz/loop.json for mode (manual/auto), enforces max_iterations and time_budget_hours guards, coordinates all components; (2) Huginn hypothesis generator (spawnHuginn) writes structured findings to .kenaz/huginn-findings.json; (3) Ansuz lifecycle delegate (spawnAnsuz via buildAnsuzPrompt) creates tasks, audits, dispatches, and self-repairs; (4) Heimdall measurement (triggerGuardianHealthCheck + save_health_snapshot) emits KEEP/REVERT verdict; (5) Wishing Well human input channel (wishes_router + frontend API); (6) Dispatch Lock Guard (isProcessAlive + launch lock section in headless.js) — per-project atomic lock file with PID liveness check prevents parallel dispatch on same working directory. Three production failure modes fixed (2026-03-23): stale locks (PID check), premature stop (countRemainingHighPriorityTasks), duplicate dispatch (prompt-level anti-duplicate in buildAnsuzPrompt). Heartbeat system (added post 2026-03-23): loop.js writes heartbeat every 10s to .kenaz/loop-heartbeat.json via startHeartbeat()/stopHeartbeat()/writeHeartbeat(); heartbeat path passed to Huginn and Ansuz sub-processes via --loop-heartbeat=HEARTBEAT_PATH flag for child process lifecycle tracking; heartbeatTimer uses .unref() so it does not prevent Node.js exit. Loop config in .kenaz/loop.json, logs in .kenaz/logs/loop/ and .kenaz/loop-log.jsonl.
-
-## Behavior
-- 先讀再改，理解現有結構後才動手
-- 遵循 codebase 現有風格，不引入新慣例
-- 只改必要的部分，不順便重構
-- 發現超出範圍的問題，在摘要中回報，不擅自處理
-- 如有測試，執行確認不破壞現有功能
-
-## Task
-**Task ID**: TASK_1818
-**Title**: post-complete 把使用者的 dispatch 設定當成 daemon 暫態狀態還原掉：dashboard-models.json 被列入 daemon-manifest state_files，每次派工後 git checkout HEAD 覆蓋使用者選的模型
-**Task File**: C:\Users\user\Documents\GitHub\KenazAI\.kenaz\tasks\active\TASK_1818_post_complete_把使用者的_dispatch_設定當成_daemon_xyz951610.yaml
-**Description**: 來源：外部 bug 回報 formation-dispatch-issues-20260726.md 的問題 6（子專案 Ashfall，另一台機器）。回報者已重現，本地已由主線程獨立定位到確切根因 —— 且根因與回報者的推論不同，其建議修法會造成新的問題。
-
-回報者觀察（可信，證據充分）：
-派工前 .kenaz/dashboard-models.json 為 dispatchFamily=bifrost / dispatchAdapter=openai，派工確實以該設定送出（回傳 adapter=openai, adapterSource=dashboard-models），但 post-complete merge 之後該檔變回 dispatchModel=claude-fable-5[1m] / dispatchFamily=claude / dispatchAdapter=null，且 updatedAt 回到舊值 2026-07-18T19:15:41.688Z。是「還原舊副本」而非重新寫入。被還原的共三個檔案，全在 .kenaz/ 底下：dashboard-models.json、specialist/turn_context_state.json、transcripts/index.json；Assets/ 底下五個他人未提交的修改全部存活。
-
-回報者的推論（不正確）：merge 流程對工作樹做破壞性操作，PRESERVE_WORKING_TREE 白名單只有一個條目所以其餘 .kenaz/ 檔案兩邊都不覆蓋而遺失。建議修法是「保留 .kenaz/ 下所有未提交修改」。
-
-實際根因（主線程讀碼定位）：
-.kenaz/specialist/daemon-manifest.yaml 的 state_files 區段明確列出 .kenaz/dashboard-models.json（該檔第 46 行附近，與 transcripts/index.json、specialist/turn_context_state.json 相鄰）。git-merge.js 的 revertDaemonManagedFiles（約 605-619 行）對清單中每個路徑執行 git checkout HEAD -- "<path>"，並印出「Reverted N daemon-managed file(s) to HEAD」。
-
-也就是說這不是覆蓋不到，是**刻意還原**。三個被還原的檔案剛好就是 manifest 裡的三個；Assets/ 存活是因為不在 manifest 裡。觀察與機制完全吻合。
-
-真正的缺陷是分類錯誤：dashboard-models.json 不是 daemon 寫的暫態狀態，它是使用者透過 dashboard UI 選擇模型時編輯的設定檔，經 dashboard_settings.rs 的 post_models 寫入。把使用者設定檔放進「派工後一律還原成 HEAD」的清單，等於每次派工都靜默丟棄使用者的選擇。
-
-為什麼回報者的建議修法不能採用：
-daemon-manifest 的存在有明確理由，檔頭寫著「These must be excluded from task commits and reverted before git merge operations to prevent conflicts」，清單內註解記載 2026-05-01 因觀察到 daemon-drift 而建、TASK_1416 因 handoff snapshot 造成 unmerged_index 而擴充。若改成「保留 .kenaz/ 下所有未提交修改」，等於停用整個 daemon-drift 防護，會讓 TASK_1416 修掉的 unmerged_index 回歸。
-
-正確的方向在 manifest 檔頭自己的措辭裡：它同時承擔兩件可分離的職責 —— (a) 排除在任務提交之外，(b) merge 前還原成 HEAD。dashboard-models.json 需要 (a)，不需要 (b)。目前的資料結構把兩者綁死成同一份清單，這才是要修的地方。
-
-嚴重性理由（回報者的三點成立）：
-1. 還原後的值指向 claude-fable-5，回報者說那正是先前秒死於額度耗盡的模型。使用者以為自己設了 GPT，下次派工卻打到一個已知會失敗的模型，且完全無提示。
-2. 這不是「設定沒存檔」，是使用者的未提交修改被無聲丟棄。
-3. 只在 session 開頭讀一次設定無法防禦。
-
-附帶事實，供評估但不必然在本卡處理：回報者稱 post-complete 全數混淆、無法逐行閱讀。在 KenazAI 原始碼倉庫中為假 —— git-merge.js 是完全可讀且註解詳盡的原始碼。回報者的環境是 D:\...\BLOCK ACTION\kenaz-core 且自述「非 git repo」，讀的是打包出貨的產物。一個會對使用者工作樹執行 git checkout 的元件以不可讀形式出貨，使得這類缺陷只能靠事後比對快照發現，這是獨立議題。
-**Priority**: P1
-**Complexity**: medium
-**Role**: developer
-
-**Scope Paths** (你只能操作這些路徑):
-- .kenaz/specialist/daemon-manifest.yaml
-- .formation/tools/post-complete/git-merge.js
-
-**Requirements**:
-- 讓「排除在任務提交之外」與「merge 前還原成 HEAD」成為兩個可分離的分類，而非共用同一份 state_files 清單
-- dashboard-models.json 改成「不進任務提交、但也不被還原」；使用者在 dashboard 選的模型必須存活過一次完整的 post-complete
-- 檢查 state_files 清單中是否還有其他使用者可編輯的設定檔被誤分類，逐項判斷並在筆記中列出判斷結果（不要只改 dashboard-models.json 就收工）
-- revertDaemonManagedFiles 還原任何檔案時必須留下可追溯的紀錄，至少列出被還原的路徑而非只印數量；目前只印 Reverted N file(s)，使用者無法從日誌得知自己的哪一個檔被丟
-- 不得採用回報者建議的「保留 .kenaz/ 下所有未提交修改」—— 那會停用 daemon-drift 防護並讓 TASK_1416 修掉的 unmerged_index 回歸
-- 保留 PRESERVE_WORKING_TREE 對 ansuz_session_handoff.json 的現有行為（TASK_1627 live-doc 契約），不要在本卡重構它
-
-**Acceptance Criteria**:
-- 實測：手動把 dashboard-models.json 改成一個與 HEAD 不同的值，跑完一次完整的 post-complete（含 merge），該值仍存活且 updatedAt 未回退
-- 實測：daemon 寫的真正暫態檔（例如 turn_context_state.json）仍然會被正確還原，防護未因本卡而失效
-- dashboard-models.json 仍不會被掃進任務 commit（以一次實跑的 commit 檔案清單為證）
-- 還原動作的日誌列出具體路徑，而非只有數量
-- 筆記列出 state_files 清單中每一項是否為使用者可編輯設定檔的逐項判斷
-- CARGO_TARGET_DIR=C:/t/mainverify node dev/scripts/run-rust-tests.js 全綠（BARE 呼叫，讀 N passed 數字，基線 1356）；若本卡未動 Rust 則註明不適用
-- 現有 post-complete 相關測試（task1536.test.js、task1543.test.js、worktree-rescue.test.js）仍綠
-
-**Technical Notes**:
-- 根因精確位置：.kenaz/specialist/daemon-manifest.yaml state_files 區段列有 .kenaz/dashboard-models.json（約第 46 行，與 .kenaz/transcripts/index.json 相鄰）；git-merge.js 約 605-619 行 revertDaemonManagedFiles 對每個路徑執行 git checkout HEAD -- "<path>"，呼叮點在 619 行 revertDaemonManagedFiles(projectRoot, '0.45')。
-- manifest 檔頭明寫「No code change required — post-complete.js reads this manifest at runtime」。若新增分類需要改 code，該句註解也要一併更新，否則它本身變成下一個講錯話的文件。
-- PRESERVE_WORKING_TREE（git-merge.js 約 29-31 行）的註解明訂「Do NOT generalize to other daemon files here — use daemon-manifest.yaml for non-live-doc daemon exclusions」。本卡的修法應該落在 manifest 側，而不是把 dashboard-models.json 塞進 PRESERVE_WORKING_TREE —— 後者正是註解明文禁止的做法。
-- git checkout HEAD -- <path> 靜默吞掉未提交修改是本 repo 反覆出現的陷阱（本 session 合併任務分支時也遇到 scoped checkout 會自動暫存的相關行為）。修改這區時先確認你理解該指令對工作樹與 index 兩者的影響。
-- 本卡改的是會對使用者工作樹執行破壞性操作的程式碼。驗證必須在可拋棄的環境或先做完整快照，不要拿使用者的主工作樹做第一次實驗。
-
-## Workflow
-### 認知 recall
-Note: 跳過此步 = 可能重複踩坑。Synapse 是預設搜尋工具，Grep 只用於精確字串匹配。
-- 用 ToolSearch 確認 synapse MCP tools 可用
-- 對 scope_paths 涉及的 symbol 呼叫 synapse_recall，了解已知的 risk/decision/fix
-- 如有相關 rune，在執行時避免踩已知地雷或破壞已記錄的設計意圖
-
-### 執行任務
-- 根據 scope_paths 讀取相關檔案
-- 根據 requirements 逐一實作
-- 根據 acceptance_criteria 驗證成果
-
-### 提交變更
-Command: git add -A && (git commit --no-verify -m "[TASK_1818] task completed" || true)
-Note: worktree 內顯式提交確保 post-complete 只需 merge 不需 recovery commit。
-nothing to commit 時靜默跳過，不報錯。
-⚠️ 若前一步有 Edit 但 git add -A 找不到變更，代表路徑用錯了（用了主 repo 路徑）。
-修復方式：先 echo $PWD 確認 worktree root，再用 worktree 絕對路徑重新 Edit。
-
-
-### 認知 inscribe
-Note: 你的發現會在對話結束時消失。Inscribe 讓未來 agent 不必重新探索。
-- 回顧工作中是否發現 non-obvious 的知識（bug 根因、隱藏約束、設計決策理由）
-- 如有，用 synapse_inscribe 記錄（confidence >= 0.6，英文撰寫）
-- 如發現 3+ symbol 構成未命名的概念，考慮用 synapse_forge 建立 Sigil
-- inscribe 後，用 synapse_rune_conflicts({ scope: '<你改的檔案路徑>' }) 檢查是否與已有 rune 矛盾
-
-### 完成任務
-Note: ⚠️ Worktree mode — 禁止執行 task-api.js complete。主 thread 在 merge + 驗證後處理。
-
-### On Blocked
-Command: node .collaboration/core/task-api.js move TASK_1818 blocked
-Note: 遇到無法解決的阻礙時，移至 blocked/ 並在摘要中說明原因
-
-## Expected Output
-Primary: 程式碼變更或檔案修改
-Secondary: 完成摘要（做了什麼、改了哪些檔案、需要注意的事）
-
-## Progress Note (mandatory — write before finishing)
-
-When you finish your work, **before your final response**, write a structured progress note to the chart.
-
-**Chart path (for reference only — do NOT use the Write tool on this path)**: `C:/Users/user/Documents/GitHub/KenazAI/.specialist/sessions/chart_TASK_1818.json`
-
-**Steps** (avoids cross-workspace permission prompts — works for same-repo and cross-project workers):
-1. Write the note JSON to a **project-local** file inside your own workspace (safe — no cross-boundary write):
-   File: `.kenaz/specialist/progress_note_TASK_1818.json`
-   Use the Write tool to create that file (it is inside your cwd, so no permission prompt occurs).
-   Ensure `.kenaz/specialist/` directory exists first (create it if absent).
-
-2. Run the phase-gate note subcommand via **Bash** to merge it into the core chart:
-   ```bash
-   node "C:/Users/user/Documents/GitHub/KenazAI/.specialist/tools/phase-gate.js" note --task=TASK_1818 --from-file=".kenaz/specialist/progress_note_TASK_1818.json"
-   ```
-   *(The note file is in your local workspace — no cross-workspace Write; phase-gate note merges it into the core chart atomically)*
-
-**Progress Note Schema** (ONE entry to append to `progress_notes`):
-
-```json
-{
-  "task_id": "TASK_1818",
-  "timestamp": "<ISO-8601>",
-  "changes": { "files_modified": ["<paths>"], "files_created": [], "files_deleted": [], "total_lines": 0 },
-  "acceptance_criteria": [
-    { "criterion": "實測：手動把 dashboard-models.json 改成一個與 HEAD 不同的值，跑完一次完整的 post-complete（含 merge），該值仍存活且 updatedAt 未回退", "status": "pass|fail|partial", "evidence": "<brief>" },
-    { "criterion": "實測：daemon 寫的真正暫態檔（例如 turn_context_state.json）仍然會被正確還原，防護未因本卡而失效", "status": "pass|fail|partial", "evidence": "<brief>" },
-    { "criterion": "dashboard-models.json 仍不會被掃進任務 commit（以一次實跑的 commit 檔案清單為證）", "status": "pass|fail|partial", "evidence": "<brief>" },
-    { "criterion": "還原動作的日誌列出具體路徑，而非只有數量", "status": "pass|fail|partial", "evidence": "<brief>" },
-    { "criterion": "筆記列出 state_files 清單中每一項是否為使用者可編輯設定檔的逐項判斷", "status": "pass|fail|partial", "evidence": "<brief>" },
-    { "criterion": "CARGO_TARGET_DIR=C:/t/mainverify node dev/scripts/run-rust-tests.js 全綠（BARE 呼叫，讀 N passed 數字，基線 1356）；若本卡未動 Rust 則註明不適用", "status": "pass|fail|partial", "evidence": "<brief>" },
-    { "criterion": "現有 post-complete 相關測試（task1536.test.js、task1543.test.js、worktree-rescue.test.js）仍綠", "status": "pass|fail|partial", "evidence": "<brief>" },
-  ],
-  "findings": ["<non-obvious discoveries, traps, design decisions>"],
-  "deviations": ["<anything done differently from the plan; REQUIRED to document a skipped inscribe>"],
-  "cognitive_output": { "runes": ["<rune_ids -- [] if none>"], "sigils": ["<sigil_ids -- [] if none>"] }
-}
-```
-
-**Rules**:
-- Evaluate EVERY Acceptance Criterion; use exact criterion text from the task.
-- `changes`: list ALL files touched (created/modified/deleted).
-- `findings`: non-obvious constraints, design traps, discoveries -- [] if none.
-- `cognitive_output.runes/sigils` = IDs from synapse calls ([] if none); if you skip inscribe, say why in `deviations` (verify-identity reads it as the skip explanation).
-- If the chart file does not exist, the note subcommand creates it automatically -- NEVER use the Write tool on the chart path directly.
-
-
-## Before You Start — Cognitive Recall (mandatory verdict)
-Before writing any code, you MUST perform a cognitive recall (synapse_recall or synapse_pulse) on the symbols you are about to modify.
-
-Output exactly one of these lines before you start coding:
-
-```
-📡 Cognitive Recall: {one-line summary of what pulse/recall revealed}
-```
-
-Or, if you consciously decide to skip:
-
-```
-📡 Cognitive Recall: SKIP — {reason why recall is unnecessary}
-```
-
-You may SKIP, but you must say why. Silence is not allowed.
-
-## Before You Finish — Don't Let Your Discovery Die
-You just spent real effort understanding this code. When this conversation ends, everything you learned disappears — unless you write it down.
-
-Ask yourself:
-- Did you trace a bug to a root cause that wasn't obvious from the code? → The next agent will waste hours rediscovering it. **inscribe type=fix**
-- Did you find a function that behaves differently than its name suggests, or has a hidden constraint? → That's a trap waiting to spring. **inscribe type=risk**
-- Did you choose approach A over B for a reason that isn't obvious? → Without your reasoning, someone will undo your work. **inscribe type=decision**
-- Did you realize that 3+ symbols are actually part of the same concept? → Give it a name. **synapse_forge** a Sigil.
-
-If none of these apply, skip. Don't write for the sake of writing.
-
-When you do write:
-- **English only** — code is English, runes should be too. This is mandatory: all memory write fields (synapse_inscribe content/context, hippocampus episode scope/lesson/summary) MUST be in English. SQLite FTS5 tokenizer does not handle CJK — Chinese content will return 0 results on recall. (Rule: MEM-LANG-001)
-- confidence >= 0.6 — if you're guessing, don't inscribe.
-- One good rune > three mediocre ones.
-- ⚠️ **Always pass `projectPath` to all Synapse tool calls** — use the projectPath determined in Phase 0. Multi-session collisions occur when concurrent sessions omit projectPath and fall back to the global default project.
-
-Every rune you inscribe will be recalled by the next agent who touches the same symbol. Your insight becomes their head start.
-
-## Cognitive Checkpoint (mandatory — include in your final output)
-At the very end of your response, output this block exactly:
-
-```
-## Cognitive Checkpoint
-Pre-code: RECALLED — {what you found} | or SKIP — {reason}
-Post-code: INSCRIBED — {rune_id} | or SKIP — {reason}
-```
-
-Both verdicts are required. You may SKIP either, but you must state why.
-
-
-## Cortex Context (Related Files Preview)
-The following files are in your scope. Review them before making changes.
-
-### .kenaz/specialist/daemon-manifest.yaml (79 lines)
-```
-# Daemon-Managed Paths Manifest
-# Files and directories written by background daemons (Ansuz, Muninn, Thalamus, etc.)
-# during active sessions. These must be excluded from task commits and reverted
-# before git merge operations to prevent conflicts.
-#
-# To add a new daemon-managed path: add it to the appropriate list below.
-# No code change required — post-complete.js reads this manifest at runtime.
-#
-# See: structural audit SECTION D.2 in post-complete.js
-
-# Files written by specialist daemons during active sessions
-state_files:
-  - .kenaz/specialist/turn_context_state.json
-  - .kenaz/specialist/ansuz_rules.yaml
-  - .kenaz/specialist/ansuz_session.json
-  - .kenaz/specialist/ansuz_state.json
-  - .kenaz/specialist/ansuz_inbox.json
-  - .kenaz/specialist/ansuz_predictions.json
-  - .kenaz/specialist/ansuz_skill_matrix.json
-  - .kenaz/specialist/attention_buffer.json
-  - .kenaz/specialist/cognitive_health.json
-  - .kenaz/specialist/daemon_outbox.json
-  - .kenaz/specialist/muninn_observations.json
-  - .kenaz/specialist/ambient_signals.json
-  - .kenaz/huginn-attention.json
-  - .kenaz/specialist/ansuz_ritual_log.jsonl
-  - .kenaz/specialist/ritual_state.json
-  - .kenaz/specialist/turn_feedback.jsonl
-  - .kenaz/specialist/ansuz_motivational_state.json
-  - .kenaz/specialist/ansuz_self_model.json
-... (truncated)
-```
-
-### .formation/tools/post-complete/git-merge.js (2329 lines)
-```
-/**
- * GIT_MERGE — extracted from post-complete.js (D.4 refactor)
- *
- * Git branch management flow: task branch creation, commit, merge to user
- * branch. Supports both direct mode (gitBranchAndMerge) and worktree-isolated
- * mode (_gitMergeFromWorktree) with rebase-under-lock, conflict detection,
- * salvage branch preservation, and daemon-managed path handling.
- */
-
-const path = require('path');
-const fs = require('fs');
-const { execSync, spawnSync } = require('child_process');
-const { ROOT, COLLABORATION_API } = require('../utils');
-const { verify: verifyAcceptance } = require('../verify-acceptance');
-// TASK_1530: salvageWorkingTreePaths extracted to a shared module so runner.js's
-// worktree-leak detector can reuse the exact same plumbing (no duplicate impl).
-// TASK_1785: readDispatchDirtySnapshot — pre-dispatch dirty snapshot so the rescue
-// scan never sweeps the user's PRE-EXISTING WIP into the worktree/merge.
-const { salvageWorkingTreePaths, readDispatchDirtySnapshot } = require('../git-salvage');
-
-// TASK_1627: Files that are daemon-managed (reverted by revertDaemonManagedFiles)
-// but are ALSO live-doc authored by the main thread (HANDOFF LIVE DOC mandate).
-// These must be saved before step-0.45 revert and restored after the merge flow
-// on ALL exit paths (try/finally). The revert keeps the merge window clean; the
-// restore ensures the next session boot inherits the most recent working-tree state,
-// not the last committed version. Do NOT generalize to other daemon files here —
-// use daemon-manifest.yaml for non-live-doc daemon exclusions. If a second live-doc
-// appears, extend this array; do NOT remove the comment explaining why.
-const PRESERVE_WORKING_TREE = [
-  '.kenaz/specialist/ansuz_session_handoff.json', // live-doc: updated each session turn
-... (truncated)
-```