maestro-flow 0.5.36 → 0.5.37

package/.claude/commands/odyssey-debug.md

@@ -1,479 +1,479 @@
----
-name: odyssey-debug
-description: Long-running debug cycle — archaeology, diagnosis, fix, confirmation, generalization, discovery, and knowledge persistence
-argument-hint: "<issue> [--skip-fix] [--skip-generalize] [--auto] [-y] [-c]"
-allowed-tools:
-  - Read
-  - Write
-  - Edit
-  - Bash
-  - Glob
-  - Grep
-  - Agent
-  - AskUserQuestion
----
-<purpose>
-Closed-loop deep debugging: archaeology → explore → diagnose → fix & confirm → generalize → discover siblings → persist.
-Treats every bug as a learning signal with exhaustive iteration until root cause confirmed or INCONCLUSIVE.
-</purpose>
-
-<boundary>
-**范围内:** 单一 bug/issue 的完整闭环 — 考古 → 探索 → 诊断 → 修复 → 确认 → 泛化同类 → 沉淀
-**范围外:** 新功能开发 → `/odyssey-planex` | 代码质量审查 → `/odyssey-review-test-fix` | UI 视觉优化 → `/odyssey-ui` | 架构重设计 → `/maestro-plan`
-**探索自由度:** 边界内自由探索 — 可追踪任意调用链、分析任意历史、测试任意假设。泛化阶段可扫描全项目寻找同类问题。
-**Zero-residual principle:** Every finding MUST have a concrete action (fix / issue / decision). "Report and shelve" is not allowed. "Pre-existing issue" is not a valid skip reason — if discovered within scope, it must be addressed.
-**模板支持:** `--template <name>` 从预定义调查策略启动，跳过假设生成直接进入针对性诊断：
-
-| Template | 调查策略 | 适用场景 |
-|----------|---------|---------|
-| `performance` | profiling → hot path → allocation → cache | 性能劣化 |
-| `memory-leak` | heap snapshot → retention chain → lifecycle | 内存泄漏 |
-| `race-condition` | timeline → concurrent access → lock analysis | 竞态条件 |
-| `regression` | git bisect → diff analysis → boundary check | 回归问题 |
-| `crash` | stack trace → null chain → error propagation | 崩溃/异常 |
-</boundary>
-
-<execution_discipline>
-**三条铁律（所有阶段适用）:**
-
-1. **Phase auto-commit** — 每个阶段完成后**自动** `git commit`，无需用户确认
-   - 代码变更 + understanding.md → `git add` → `git commit -m "odyssey-debug({slug}): {phase} — {摘要}"`
-   - session.json / evidence.ndjson 为运行时状态，不纳入 commit
-   - 确保每个阶段的进展可回溯、可恢复
-
-2. **Confident edits only, but must attempt** — only modify what you're confident about; record decisions only when genuinely requiring human judgment
-   - Confident → edit code directly, commit
-   - Needs decision → record `evidence.ndjson {"phase":"decision","status":"pending"}`, don't touch code
-   - No speculative changes
-   - ⚠️ **Decision gate** — ONLY these qualify as decisions (not fixes):
-     - Cross-module architectural tradeoffs requiring human direction
-     - Ambiguous business semantics where the fix could alter intended behavior
-     - Requires new dependency or breaking API change
-   - ❌ "Unsure how to fix", "Large scope", "Pre-existing issue" are NOT valid decision reasons — either fix it, or explain specifically why it's unfixable
-
-3. **多 CLI 辅助** — 利用 `maestro delegate` 调用多个 CLI 工具交叉验证
-   - 关键判断用不同 `--role`（analyze / review / explore）获取多视角
-   - 修复前后各做一次 CLI review 确认
-   - 不同阶段可调用不同工具，综合多方意见再行动
-
-4. **禁止以上下文消耗为由中断** — harness 自动处理 context compression，以"上下文不足"或"已执行 N 个阶段"为由中断属于纪律违反；必须完整走完状态机直到 S_RECORD → END
-</execution_discipline>
-
-<context>
-$ARGUMENTS — issue description and optional flags.
-
-**Flags:** `--skip-fix` analysis-only | `--skip-generalize` quick fix | `--template <name>` 预定义策略 | `--auto` no delegate confirmation | `-y` auto-confirm all decisions | `-c` resume last session
-
-**Session**: `SESSION_DIR = .workflow/scratch/{YYYYMMDD}-debug-odyssey-{slug}/`
-
-**Output — 4 files:**
-```
-SESSION_DIR/
-  ├── session.json       # state + confirmation + patterns + phase_goals
-  ├── evidence.ndjson    # append-only evidence trail (phase field distinguishes origin)
-  ├── explore.json       # structured CLI exploration snapshot
-  └── understanding.md   # evolving narrative — 9 sections
-```
-
-**session.json schema:**
-```json
-{
-  "session_id": "debug-odyssey-{YYYYMMDD-HHmmss}", "issue": "",
-  "flags": { "skip_fix": false, "skip_generalize": false, "auto": false, "auto_confirm": false },
-  "current_state": "S_INTAKE", "diagnosis_retries": 0,
-  "root_cause": null, "patterns": [], "confirmation": null,
-  "phase_goals": [], "phase_goals_all_done": false, "self_iteration_log": [],
-  "generalization_stats": null,
-  "cross_phase_loops": 0, "max_loops": 5,
-  "created_at": "", "updated_at": ""
-}
-```
-
-**evidence.ndjson — unified trail:**
-```json
-{"ts":"","phase":"archaeology|explore|diagnosis|discovery|decision|self-iteration","type":"","source":"","content":"","note":""}
-```
-Phase-specific fields:
-- `archaeology`: `sha`, `author`, `date`, `message`, `relevance` (high|medium|low)
-- `explore`: `category` (call_chain|recent_change|error_gap|similar_pattern), `detail`
-- `diagnosis`: `hypothesis`, `result` (confirmed|disproved|inconclusive)
-- `discovery`: `file`, `line`, `classification` (safe|risk|bug), `action` (fix|issue|decision|skip)
-- `decision`: `question`, `options`, `context`, `status` (pending|resolved|deferred), `resolution`
-- `self-iteration`: `stage`, `round`, `assessment`, `expansion`
-
-**explore.json schema:**
-```json
-{
-  "call_chains": [{"entry":"","chain":["file:line"]}],
-  "recent_changes": [{"file":"","commits":[{"sha":"","message":"","date":""}]}],
-  "error_gaps": [{"file":"","line":0,"description":""}],
-  "similar_patterns": [{"file":"","line":0,"description":""}],
-  "cli_tool": "", "timestamp": ""
-}
-```
-
-**phase_goals[] — auto-derived from flags:**
-
-| ID | Goal | done_when | phase | skip_when |
-|----|------|-----------|-------|-----------|
-| G1 | Root cause identified | evidence.ndjson has phase=diagnosis result=confirmed | S_DIAGNOSE | — |
-| G2 | Explore context gathered | explore.json ≥1 category populated | S_EXPLORE | — |
-| G3 | Fix applied and confirmed | confirmation.overall == confirmed | S_CONFIRM | skip_fix |
-| G4 | Pattern generalized | patterns[] ≥1 entry | S_GENERALIZE | skip_generalize |
-| G5 | Discoveries triaged | all scan hits classified | S_DISCOVER | skip_generalize |
-| G6 | Learnings persisted | spec entries created OR no actionable learnings | S_RECORD | — |
-
-When `flags[skip_when] == true` → auto set `status: "skipped"`, `completion_confirmed: true`.
-
-**understanding.md — 9 sections (written by owning phase):**
-1. Issue & Scope ← S_INTAKE | 2. Archaeology Summary ← S_ARCHAEOLOGY | 3. Exploration ← S_EXPLORE
-4. Hypotheses & Testing ← S_DIAGNOSE | 5. Root Cause ← S_DIAGNOSE | 6. Fix & Confirmation ← S_FIX+S_CONFIRM
-7. Generalization ← S_GENERALIZE | 8. Discoveries & Decisions ← S_DISCOVER | 9. Learnings ← S_RECORD
-
-### Pre-load（可选，缺失不阻塞）
-
-| 层级 | 命令 | 作用 |
-|------|------|------|
-| Codebase docs | Read `.workflow/codebase/ARCHITECTURE.md` | 模块边界，作为所有分析的上下文 |
-| Wiki search | `maestro search "<issue keywords>" --json` | 先前调查、相关决策（取 top 5） |
-| Specs + tools | `maestro spec load --category debug --keyword "<symptom>"` | 已知 issue/workaround + 可发现的 knowhow 工具 |
-| Role knowledge | `maestro search --category debug` → 选相关 → `maestro wiki load <id>` | 累积领域知识 |
-| Prior sessions | `Glob(".workflow/scratch/*-debug-odyssey-*")` | 相关 odyssey 会话 |
-
-### Knowledge Persistence（S_RECORD 中写入产出文件）
-
-S_RECORD 阶段将可沉淀知识 **写入 understanding.md §9 Learnings**，按以下分类结构化：
-
-| 分类 | 写入内容 | 后续建议命令 |
-|------|---------|-------------|
-| 反复根因模式 | 模式描述 + 触发条件 + 修复模板 | `/spec-add debug "..."` |
-| 非显而易见 workaround | 问题场景 + 解决方案 + 适用范围 | `/spec-add learning "..."` |
-| 架构边界违反 | 违反描述 + 正确边界 + 检查方法 | `/spec-add arch "..."` |
-| 可复用泛化 pattern | pattern 签名 + 风险说明 + fix 模板 | `/spec-add coding "..."` |
-
-**两步模式：** 执行中写入产出文件（临时记录）→ 任务完成后用户通过 next_step_routing 沉淀为永久知识。执行过程中不调用外部 Skill。
-</context>
-
-<self_iteration>
-**Quality Gate (适用: S_ARCHAEOLOGY, S_EXPLORE, S_DIAGNOSE, S_GENERALIZE)**
-
-| 维度 | sufficient | insufficient |
-|------|-----------|-------------|
-| Coverage | 已知相关文件/模块均已分析 | 遗漏 grep/git log 可发现的目标 |
-| Depth | ≥80% 发现有 file:line 级证据 | 多数仅泛泛描述 |
-| Actionability | 每条结论有具体后续动作 | 仅"建议关注"类无操作性结论 |
-
-**Rules:** phase complete → evaluate 3 dimensions → any insufficient → re-enter (max **3 rounds** per stage).
-- Round 1: widen scope — more dirs, git log depth ×2, extra delegate angles
-- Round 2: shift perspective — different CLI tool, reverse trace, manual code reading
-- Round 3: combine both + targeted deep-dive on remaining gaps
-
-**Exit:** all sufficient → advance | 3-round cap → log remaining gaps and continue. Record to `evidence.ndjson` + `session.json.self_iteration_log[]`.
-</self_iteration>
-
-<state_machine>
-
-<states>
-S_INTAKE       — 解析问题、加载上下文、检查/恢复 session     PERSIST: session.json + understanding.md §1
-S_ARCHAEOLOGY  — 考古：git history + CLI 分析              PERSIST: evidence.ndjson (archaeology) + understanding.md §2
-S_EXPLORE      — CLI 探索：调用链、错误间隙、相似模式        PERSIST: explore.json + evidence.ndjson (explore) + understanding.md §3
-S_DIAGNOSE     — 假设驱动根因分析                          PERSIST: evidence.ndjson (diagnosis|decision) + understanding.md §4-5
-S_FIX          — 实现修复 (skip_fix 时跳过)                PERSIST: code changes + evidence.ndjson (decision)
-S_CONFIRM      — 测试 + CLI review 双重确认 (skip_fix 时跳过) PERSIST: session.json.confirmation + understanding.md §6
-S_GENERALIZE   — 举一反三：提取 pattern，扫描相似代码       PERSIST: session.json.patterns + understanding.md §7
-S_DISCOVER     — 评估发现，创建 issue / 记录决策            PERSIST: evidence.ndjson (discovery|decision) + understanding.md §8
-S_RECORD       — 知识沉淀 + 目标审计                       PERSIST: understanding.md §9 + spec entries
-</states>
-
-<transitions>
-S_INTAKE → S_INTAKE       : -c + session found → A_RESUME_SESSION
-S_INTAKE → S_ARCHAEOLOGY  : issue parsed → A_INTAKE
-S_INTAKE → S_INTAKE       : no issue, no session → AskUserQuestion
-
-S_ARCHAEOLOGY → S_EXPLORE     : A_ARCHAEOLOGY complete
-S_EXPLORE     → S_DIAGNOSE    : A_EXPLORE complete
-
-S_DIAGNOSE → S_FIX          : root cause confirmed, !skip_fix
-S_DIAGNOSE → S_GENERALIZE   : root cause confirmed, skip_fix, !skip_generalize
-S_DIAGNOSE → S_RECORD       : root cause confirmed, skip_fix, skip_generalize
-S_DIAGNOSE → S_DIAGNOSE     : all hypotheses failed, retries < 3 → A_ESCALATE_DIAGNOSIS
-S_DIAGNOSE → S_RECORD       : all hypotheses failed, retries >= 3 → mark INCONCLUSIVE
-
-S_FIX     → S_CONFIRM       : fix implemented
-S_CONFIRM → S_GENERALIZE    : confirmed, !skip_generalize
-S_CONFIRM → S_RECORD        : confirmed, skip_generalize
-S_CONFIRM → S_FIX           : needs_rework
-
-S_GENERALIZE → S_DISCOVER   : similar code found
-S_GENERALIZE → S_RECORD     : no similar code
-
-S_DISCOVER → S_DIAGNOSE     : discovery finds new bug worth investigating → cross_phase_loops++
-S_DISCOVER → S_FIX          : discovery finds same-pattern bug, fix template applies, !skip_fix → cross_phase_loops++
-S_DISCOVER → S_RECORD       : triage complete AND remaining_actionable == 0
-S_DISCOVER → S_RECORD       : loops >= max_loops → MUST log each unfixed item with specific reason (blanket "pre-existing" is forbidden)
-
-S_RECORD   → END            : A_RECORD complete
-</transitions>
-
-<actions>
-
-### A_INTAKE
-1. Parse arguments: issue description, flags
-2. Generate slug, create `SESSION_DIR`
-3. Search: `maestro search "<keywords>"` + Glob prior sessions + ARCHITECTURE.md + Grep keywords
-4. Derive `phase_goals[]` from flags (apply `skip_when`)
-5. Write `session.json` + `understanding.md` §1
-6. Emit Goal Prompt (see Appendix)
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): INTAKE — 目标解析与上下文加载"`
-
-### A_RESUME_SESSION
-Find latest session via Glob → read `session.json` → display summary → jump to `current_state`.
-
-### A_ARCHAEOLOGY
-**Git archaeology (2 parallel Agents):**
-
-| Agent | Task |
-|-------|------|
-| Timeline | `git log --oneline -20 -- {files}` → change timeline |
-| Blame | Top 3 suspicious files: `git blame -L {region}` → critical paths |
-
-Append findings to `evidence.ndjson` (phase: "archaeology").
-
-**CLI-assisted change review:**
-```bash
-maestro delegate "PURPOSE: Review recent modifications related to: {issue}
-TASK: Analyze intent behind changes | Identify risky modifications | Flag potential bug sources
-MODE: analysis
-CONTEXT: @{relevant_files} | Git log: {top_10_commits}
-EXPECTED: JSON [{commit_sha, risk_level, analysis, could_cause_issue, explanation}]
-CONSTRAINTS: Focus on behavioral changes, not formatting
-" --role analyze --mode analysis
-```
-Execute with `run_in_background: true`, then wait for callback (do NOT halt the Odyssey flow). Append results to evidence.
-
-Update `understanding.md` §2.
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): ARCHAEOLOGY — git 考古分析"`
-
-### A_EXPLORE
-Skip if no enabled CLI tools (W006).
-
-```bash
-maestro delegate "PURPOSE: Gather codebase evidence for: {issue}
-TASK: Trace call chains | Find recent changes | Identify error gaps | Check similar patterns
-MODE: analysis
-CONTEXT: @**/*
-EXPECTED: JSON {call_chains, recent_changes, error_gaps, similar_patterns}
-CONSTRAINTS: Max 20 entries/category | Symptom-related code paths
-Symptoms: {issue}  Archaeology hints: {suspicious_commits}
-" --role explore --mode analysis
-```
-Execute with `run_in_background: true`, then wait for callback (do NOT halt the Odyssey flow).
-
-Parse → write `explore.json` + append `evidence.ndjson` (phase: "explore"). Update §3. Mark G2 done.
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): EXPLORE — 代码探索完成"`
-
-### A_DIAGNOSE
-1. **Form hypotheses** from evidence (archaeology + explore), ranked [HIGH]/[MEDIUM]/[LOW] → §4
-2. **Test each** (rank order): design test → execute → append evidence (phase: "diagnosis")
-3. **Decision journal**: ambiguity → evidence (phase: "decision"); Normal: AskUserQuestion | `-y`: defer
-4. **Root cause**: confirmed → `session.json.root_cause` + §5. Mark G1 done.
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): DIAGNOSE — 根因确认"`
-
-### A_ESCALATE_DIAGNOSIS
-Increment `diagnosis_retries`. If < 3: broaden scope via `maestro delegate --role analyze` (same delegate format), form new hypotheses, return to S_DIAGNOSE. If >= 3: Normal → AskUserQuestion (broaden/new/INCONCLUSIVE) | `-y` → auto INCONCLUSIVE, proceed to S_RECORD. See Appendix: `-y` behavior.
-
-### A_FIX
-1. Present root cause + proposed fix. Normal: AskUserQuestion | `-y`: auto proceed (see Appendix)
-2. Implement fix
-3. Record in evidence (phase: "decision")
-
-📌 **Auto-commit**: `git add -A && git commit -m "odyssey-debug({slug}): FIX — {修复摘要}"`
-
-### A_CONFIRM
-1. **Tests**: auto-detect framework, run covering tests
-2. **CLI fix review**:
-```bash
-maestro delegate "PURPOSE: Review fix for: {issue}
-TASK: Verify correctness | Check regressions | Assess completeness | Review edge cases
-MODE: analysis
-CONTEXT: @{modified_files} | Root cause: {summary} | Diff: {git_diff}
-EXPECTED: JSON {verdict, findings [{severity, description, suggestion}], regression_risk}
-CONSTRAINTS: Focus on correctness, not style
-" --role review --mode analysis
-```
-Execute with `run_in_background: true`, then wait for callback (do NOT halt the Odyssey flow).
-
-3. Write `session.json.confirmation`: `{test_result, cli_review, overall: "confirmed|needs_rework", timestamp}`
-4. Update §6. `needs_rework` → S_FIX. `confirmed` → mark G3 done, advance.
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): CONFIRM — 修复验证"`
-
-### A_GENERALIZE
-举一反三: multi-layer pattern extraction → 4-agent scan → cross-layer dedup → iterative deepening.
-
-**Pattern extraction** from root cause + fix:
-
-| Layer | Method | Example |
-|-------|--------|---------|
-| Syntax | Regex patterns (direct Grep) | `eval(`, missing `await`, unclosed resource |
-| Semantic | Anti-pattern description (Agent-driven) | Unhandled async errors, unvalidated input |
-| Structural | Architecture-level (file/module similarity) | Same import structure, missing override |
-
-Write `session.json.patterns[]`: `[{id, source, layer, signature, description, risk, fix_template}]`
-
-**4-agent parallel codebase scan:**
-
-| Agent | Strategy | Input | Scope |
-|-------|----------|-------|-------|
-| Syntax grep | Grep syntax-layer regex | P*.signature | Full project |
-| Semantic scan | Understand + check anti-pattern | P*.description | Related modules |
-| Structural match | Find structurally similar files | Buggy file structure | Full project |
-| Historical grep | `git log -S "{pattern}"` | P*.signature | Full git history |
-
-Returns: `[{pattern_id, file, line, context, risk_level, layer, confidence}]`
-
-**Cross-layer dedup**: same file:line multi-layer hit → boost confidence | single-layer → `needs_review` | historical hit on fixed record → `regression_risk`
-
-**Iterative deepening**: module with ≥3 hits → targeted deep scan (max 1 round).
-
-**Quality Gate** (self-iteration) → if insufficient, expand and re-scan.
-
-Write §7 + `session.json.generalization_stats`: `{patterns_extracted, total_hits, cross_layer_confirmed, regression_risks, by_layer, deepening_triggered}`. Mark G4 done.
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): GENERALIZE — 泛化扫描完成"`
-
-### A_DISCOVER
-1. **Triage** each hit: read ±10 lines context → classify `safe`/`risk`/`bug`
-2. **Route**:
-   - `bug` + directly fixable → **fix immediately** (not just log an issue) → back to S_FIX
-   - `bug` + requires cross-module/architectural decision → create issue (with fix suggestion + impact analysis)
-   - `risk` → evaluate if guard/validation can mitigate directly; if yes, fix it
-   - `safe` → mark skip
-   See Appendix `-y` behavior table. Append evidence (phase: "discovery" + "decision")
-3. Update §8. Mark G5 done.
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): DISCOVER — 发现分类完成"`
-
-### A_RECORD
-1. Finalize `understanding.md` §9
-2. **Write learnings** to understanding.md §9: 按 Knowledge Persistence 表分类记录（临时），completion summary 列出建议的 `/spec-add` 命令
-3. Mark G6 done. Process pending decisions: Normal → AskUserQuestion | `-y` → skip (show deferred count)
-4. **Goal audit**: all `completion_confirmed` true → `phase_goals_all_done = true`. Any false: Normal → AskUserQuestion (回退/跳过/接受) | `-y` → auto accept
-5. **Completion**: `current_state = "COMPLETED"`, emit summary:
-```
---- DEBUG ODYSSEY COMPLETE ---
-Issue:      {issue}
-Root cause: {root_cause.hypothesis}
-Fix:        {applied|skipped|inconclusive}
-Patterns:   {patterns_extracted} ({by_layer} distribution)
-Scan hits:  {total_hits} ({cross_layer_confirmed} cross-layer confirmed)
-Issues:     {N} created
-Decisions:  {N} resolved, {M} pending, {K} deferred
-Learnings:  {N} spec entries persisted
-Self-iter:  {N} quality gate rounds across {M} stages
-Goals:      {done}/{total} confirmed ({skipped} skipped)
----
-```
-
-📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): RECORD — 会话总结与知识沉淀"`
-
-</actions>
-
-<appendix>
-
-### Goal Prompt Template
-
-**时机守卫：仅在 A_INTAKE 完成后显示一次。A_RECORD 完成时禁止重新显示。**
-
-```
-📋 Debug Odyssey 会话已创建。可随时复制以下 /goal 设定终止条件：
-
-/goal 完成以下目标：
-{for each G in phase_goals where status != "skipped":}
-- {G.id}: {G.goal} — 完成条件: {G.done_when}
-{end for}
-穷尽迭代：直到根因确认（或明确 INCONCLUSIVE）且修复验证通过
-且泛化扫描穷尽（所有 hits 已分类处理）且 phase_goals_all_done=true 才停。
-假设失败时扩范围→换视角→升级工具，不轻易放弃。
-泛化发现的同类 bug 全部修复或创建 issue，不允许遗留。
-遇到 phase=decision 的 pending 必须 AskUserQuestion，不得自行 resolve。
-```
-
-Odyssey 输出提示词后继续执行不阻塞。`/goal` 由用户任意时刻输入。
-
-### `-y` Auto-Confirm Behavior
-
-| Decision Point | Normal | `-y` mode |
-|---------------|--------|-----------|
-| A_DIAGNOSE ambiguity | AskUserQuestion blocks | record `deferred`, best-effort continue |
-| A_ESCALATE 3-strike | AskUserQuestion 3-way | auto INCONCLUSIVE |
-| A_FIX direction | AskUserQuestion confirm | auto proceed with suggested fix |
-| A_DISCOVER bug triage | AskUserQuestion route | auto create issue |
-| A_DISCOVER ambiguous | AskUserQuestion batch | all `deferred` |
-| A_RECORD decisions | AskUserQuestion per-item | skip, show deferred count |
-| A_RECORD goal audit | AskUserQuestion 3-way | auto accept current state |
-
-`deferred` items shown as "待决策" in completion summary; recoverable via `-c`.
-
-### Phase Goal Lifecycle
-
-`pending → done (confirmed=true)` normal | `pending → skipped (confirmed=true)` flags/manual | `pending → failed (confirmed=false)` INCONCLUSIVE
-
-`phase_goals_all_done = true` only when ALL goals have `completion_confirmed == true`.
-
-</appendix>
-
-</state_machine>
-
-<error_codes>
-| Code | Severity | Condition | Recovery |
-|------|----------|-----------|----------|
-| E001 | error | No issue and no session to resume | Provide issue or use -c |
-| E003 | error | Resume but no session found | Start new session |
-| E004 | error | Delegate execution failed | Retry or proceed without CLI |
-| W001 | warning | No relevant git history | Proceed with limited context |
-| W002 | warning | All hypotheses inconclusive after 3 retries | INCONCLUSIVE |
-| W003 | warning | Generalization scan 0 hits | Skip discovery |
-| W004 | warning | Delegate parse failed | Use raw output |
-| W005 | warning | Pending decisions unresolved | Filter evidence.ndjson phase=decision |
-| W006 | warning | CLI exploration skipped (no tools) | Proceed without explore.json |
-</error_codes>
-
-<success_criteria>
-- [ ] Session directory created with 4 output files
-- [ ] Prior knowledge searched (maestro search + sessions + architecture)
-- [ ] Git archaeology (log + blame) + CLI change review, evidence.ndjson phase=archaeology
-- [ ] CLI exploration, explore.json written, evidence phase=explore
-- [ ] Hypotheses formed from archaeology + explore, tested and logged phase=diagnosis
-- [ ] Root cause declared with evidence refs
-- [ ] understanding.md tracks all 9 sections progressively
-- [ ] Fix implemented + confirmed with test + CLI review (unless --skip-fix)
-- [ ] Multi-layer patterns (syntax/semantic/structural) extracted (unless --skip-generalize)
-- [ ] 4-agent scan + cross-layer dedup + iterative deepening for ≥3 hits/module
-- [ ] Discoveries classified and routed (fix/issue/decision/skip)
-- [ ] **Every unfixed finding has individual classification and reason** — blanket "pre-existing" labels are forbidden
-- [ ] Decision journal: all human-judgment items in evidence.ndjson phase=decision
-- [ ] phase_goals derived from flags, skip_when applied, each phase marks its goal
-- [ ] Goal audit in A_RECORD — unmet goals surfaced, phase_goals_all_done set correctly
-- [ ] Goal Prompt displayed once after session creation
-- [ ] `-y`: all decisions auto-resolve/defer, deferred count in summary
-- [ ] State saved at each transition (resumable via -c)
-- [ ] Quality Gate self-iteration when insufficient, logged in self_iteration_log
-- [ ] Spec entries persisted for reusable learnings
-- [ ] Completion summary with all stats
-</success_criteria>
-
-<next_step_routing>
-| Condition | Next step |
-|-----------|-----------|
-| Issues from discoveries | `/manage-issue list --source debug-odyssey` |
-| Pattern worth documenting | `/learn-decompose <module>` |
-| Fix needs formal review | `/quality-review <phase>` |
-| Second opinion on root cause | `/learn-second-opinion <understanding.md>` |
-| Related question | `/learn-investigate "<question>"` |
-| Decisions still pending | Filter evidence.ndjson phase=decision status=pending |
-</next_step_routing>
+---
+name: odyssey-debug
+description: Long-running debug cycle — archaeology, diagnosis, fix, confirmation, generalization, discovery, and knowledge persistence
+argument-hint: "<issue> [--skip-fix] [--skip-generalize] [--auto] [-y] [-c]"
+allowed-tools:
+  - Read
+  - Write
+  - Edit
+  - Bash
+  - Glob
+  - Grep
+  - Agent
+  - AskUserQuestion
+---
+<purpose>
+Closed-loop deep debugging: archaeology → explore → diagnose → fix & confirm → generalize → discover siblings → persist.
+Treats every bug as a learning signal with exhaustive iteration until root cause confirmed or INCONCLUSIVE.
+</purpose>
+
+<boundary>
+**范围内:** 单一 bug/issue 的完整闭环 — 考古 → 探索 → 诊断 → 修复 → 确认 → 泛化同类 → 沉淀
+**范围外:** 新功能开发 → `/odyssey-planex` | 代码质量审查 → `/odyssey-review-test-fix` | UI 视觉优化 → `/odyssey-ui` | 架构重设计 → `/maestro-plan`
+**探索自由度:** 边界内自由探索 — 可追踪任意调用链、分析任意历史、测试任意假设。泛化阶段可扫描全项目寻找同类问题。
+**Zero-residual principle:** Every finding MUST have a concrete action (fix / issue / decision). "Report and shelve" is not allowed. "Pre-existing issue" is not a valid skip reason — if discovered within scope, it must be addressed.
+**模板支持:** `--template <name>` 从预定义调查策略启动，跳过假设生成直接进入针对性诊断：
+
+| Template | 调查策略 | 适用场景 |
+|----------|---------|---------|
+| `performance` | profiling → hot path → allocation → cache | 性能劣化 |
+| `memory-leak` | heap snapshot → retention chain → lifecycle | 内存泄漏 |
+| `race-condition` | timeline → concurrent access → lock analysis | 竞态条件 |
+| `regression` | git bisect → diff analysis → boundary check | 回归问题 |
+| `crash` | stack trace → null chain → error propagation | 崩溃/异常 |
+</boundary>
+
+<execution_discipline>
+**三条铁律（所有阶段适用）:**
+
+1. **Phase auto-commit** — 每个阶段完成后**自动** `git commit`，无需用户确认
+   - 代码变更 + understanding.md → `git add` → `git commit -m "odyssey-debug({slug}): {phase} — {摘要}"`
+   - session.json / evidence.ndjson 为运行时状态，不纳入 commit
+   - 确保每个阶段的进展可回溯、可恢复
+
+2. **Confident edits only, but must attempt** — only modify what you're confident about; record decisions only when genuinely requiring human judgment
+   - Confident → edit code directly, commit
+   - Needs decision → record `evidence.ndjson {"phase":"decision","status":"pending"}`, don't touch code
+   - No speculative changes
+   - ⚠️ **Decision gate** — ONLY these qualify as decisions (not fixes):
+     - Cross-module architectural tradeoffs requiring human direction
+     - Ambiguous business semantics where the fix could alter intended behavior
+     - Requires new dependency or breaking API change
+   - ❌ "Unsure how to fix", "Large scope", "Pre-existing issue" are NOT valid decision reasons — either fix it, or explain specifically why it's unfixable
+
+3. **多 CLI 辅助** — 利用 `maestro delegate` 调用多个 CLI 工具交叉验证
+   - 关键判断用不同 `--role`（analyze / review / explore）获取多视角
+   - 修复前后各做一次 CLI review 确认
+   - 不同阶段可调用不同工具，综合多方意见再行动
+
+4. **禁止以上下文消耗为由中断** — harness 自动处理 context compression，以"上下文不足"或"已执行 N 个阶段"为由中断属于纪律违反；必须完整走完状态机直到 S_RECORD → END
+</execution_discipline>
+
+<context>
+$ARGUMENTS — issue description and optional flags.
+
+**Flags:** `--skip-fix` analysis-only | `--skip-generalize` quick fix | `--template <name>` 预定义策略 | `--auto` no delegate confirmation | `-y` auto-confirm all decisions | `-c` resume last session
+
+**Session**: `SESSION_DIR = .workflow/scratch/{YYYYMMDD}-debug-odyssey-{slug}/`
+
+**Output — 4 files:**
+```
+SESSION_DIR/
+  ├── session.json       # state + confirmation + patterns + phase_goals
+  ├── evidence.ndjson    # append-only evidence trail (phase field distinguishes origin)
+  ├── explore.json       # structured CLI exploration snapshot
+  └── understanding.md   # evolving narrative — 9 sections
+```
+
+**session.json schema:**
+```json
+{
+  "session_id": "debug-odyssey-{YYYYMMDD-HHmmss}", "issue": "",
+  "flags": { "skip_fix": false, "skip_generalize": false, "auto": false, "auto_confirm": false },
+  "current_state": "S_INTAKE", "diagnosis_retries": 0,
+  "root_cause": null, "patterns": [], "confirmation": null,
+  "phase_goals": [], "phase_goals_all_done": false, "self_iteration_log": [],
+  "generalization_stats": null,
+  "cross_phase_loops": 0, "max_loops": 5,
+  "created_at": "", "updated_at": ""
+}
+```
+
+**evidence.ndjson — unified trail:**
+```json
+{"ts":"","phase":"archaeology|explore|diagnosis|discovery|decision|self-iteration","type":"","source":"","content":"","note":""}
+```
+Phase-specific fields:
+- `archaeology`: `sha`, `author`, `date`, `message`, `relevance` (high|medium|low)
+- `explore`: `category` (call_chain|recent_change|error_gap|similar_pattern), `detail`
+- `diagnosis`: `hypothesis`, `result` (confirmed|disproved|inconclusive)
+- `discovery`: `file`, `line`, `classification` (safe|risk|bug), `action` (fix|issue|decision|skip)
+- `decision`: `question`, `options`, `context`, `status` (pending|resolved|deferred), `resolution`
+- `self-iteration`: `stage`, `round`, `assessment`, `expansion`
+
+**explore.json schema:**
+```json
+{
+  "call_chains": [{"entry":"","chain":["file:line"]}],
+  "recent_changes": [{"file":"","commits":[{"sha":"","message":"","date":""}]}],
+  "error_gaps": [{"file":"","line":0,"description":""}],
+  "similar_patterns": [{"file":"","line":0,"description":""}],
+  "cli_tool": "", "timestamp": ""
+}
+```
+
+**phase_goals[] — auto-derived from flags:**
+
+| ID | Goal | done_when | phase | skip_when |
+|----|------|-----------|-------|-----------|
+| G1 | Root cause identified | evidence.ndjson has phase=diagnosis result=confirmed | S_DIAGNOSE | — |
+| G2 | Explore context gathered | explore.json ≥1 category populated | S_EXPLORE | — |
+| G3 | Fix applied and confirmed | confirmation.overall == confirmed | S_CONFIRM | skip_fix |
+| G4 | Pattern generalized | patterns[] ≥1 entry | S_GENERALIZE | skip_generalize |
+| G5 | Discoveries triaged | all scan hits classified | S_DISCOVER | skip_generalize |
+| G6 | Learnings persisted | spec entries created OR no actionable learnings | S_RECORD | — |
+
+When `flags[skip_when] == true` → auto set `status: "skipped"`, `completion_confirmed: true`.
+
+**understanding.md — 9 sections (written by owning phase):**
+1. Issue & Scope ← S_INTAKE | 2. Archaeology Summary ← S_ARCHAEOLOGY | 3. Exploration ← S_EXPLORE
+4. Hypotheses & Testing ← S_DIAGNOSE | 5. Root Cause ← S_DIAGNOSE | 6. Fix & Confirmation ← S_FIX+S_CONFIRM
+7. Generalization ← S_GENERALIZE | 8. Discoveries & Decisions ← S_DISCOVER | 9. Learnings ← S_RECORD
+
+### Pre-load（可选，缺失不阻塞）
+
+| 层级 | 命令 | 作用 |
+|------|------|------|
+| Codebase docs | Read `.workflow/codebase/ARCHITECTURE.md` | 模块边界，作为所有分析的上下文 |
+| Wiki search | `maestro search "<issue keywords>" --json` | 先前调查、相关决策（取 top 5） |
+| Specs + tools | `maestro load --type spec --category debug --keyword "<symptom>"` | 已知 issue/workaround + 可发现的 knowhow 工具 |
+| Role knowledge | `maestro search --category debug` → 选相关 → `maestro load --type knowhow --id <id>` | 累积领域知识 |
+| Prior sessions | `Glob(".workflow/scratch/*-debug-odyssey-*")` | 相关 odyssey 会话 |
+
+### Knowledge Persistence（S_RECORD 中写入产出文件）
+
+S_RECORD 阶段将可沉淀知识 **写入 understanding.md §9 Learnings**，按以下分类结构化：
+
+| 分类 | 写入内容 | 后续建议命令 |
+|------|---------|-------------|
+| 反复根因模式 | 模式描述 + 触发条件 + 修复模板 | `/spec-add debug "..."` |
+| 非显而易见 workaround | 问题场景 + 解决方案 + 适用范围 | `/spec-add learning "..."` |
+| 架构边界违反 | 违反描述 + 正确边界 + 检查方法 | `/spec-add arch "..."` |
+| 可复用泛化 pattern | pattern 签名 + 风险说明 + fix 模板 | `/spec-add coding "..."` |
+
+**两步模式：** 执行中写入产出文件（临时记录）→ 任务完成后用户通过 next_step_routing 沉淀为永久知识。执行过程中不调用外部 Skill。
+</context>
+
+<self_iteration>
+**Quality Gate (适用: S_ARCHAEOLOGY, S_EXPLORE, S_DIAGNOSE, S_GENERALIZE)**
+
+| 维度 | sufficient | insufficient |
+|------|-----------|-------------|
+| Coverage | 已知相关文件/模块均已分析 | 遗漏 grep/git log 可发现的目标 |
+| Depth | ≥80% 发现有 file:line 级证据 | 多数仅泛泛描述 |
+| Actionability | 每条结论有具体后续动作 | 仅"建议关注"类无操作性结论 |
+
+**Rules:** phase complete → evaluate 3 dimensions → any insufficient → re-enter (max **3 rounds** per stage).
+- Round 1: widen scope — more dirs, git log depth ×2, extra delegate angles
+- Round 2: shift perspective — different CLI tool, reverse trace, manual code reading
+- Round 3: combine both + targeted deep-dive on remaining gaps
+
+**Exit:** all sufficient → advance | 3-round cap → log remaining gaps and continue. Record to `evidence.ndjson` + `session.json.self_iteration_log[]`.
+</self_iteration>
+
+<state_machine>
+
+<states>
+S_INTAKE       — 解析问题、加载上下文、检查/恢复 session     PERSIST: session.json + understanding.md §1
+S_ARCHAEOLOGY  — 考古：git history + CLI 分析              PERSIST: evidence.ndjson (archaeology) + understanding.md §2
+S_EXPLORE      — CLI 探索：调用链、错误间隙、相似模式        PERSIST: explore.json + evidence.ndjson (explore) + understanding.md §3
+S_DIAGNOSE     — 假设驱动根因分析                          PERSIST: evidence.ndjson (diagnosis|decision) + understanding.md §4-5
+S_FIX          — 实现修复 (skip_fix 时跳过)                PERSIST: code changes + evidence.ndjson (decision)
+S_CONFIRM      — 测试 + CLI review 双重确认 (skip_fix 时跳过) PERSIST: session.json.confirmation + understanding.md §6
+S_GENERALIZE   — 举一反三：提取 pattern，扫描相似代码       PERSIST: session.json.patterns + understanding.md §7
+S_DISCOVER     — 评估发现，创建 issue / 记录决策            PERSIST: evidence.ndjson (discovery|decision) + understanding.md §8
+S_RECORD       — 知识沉淀 + 目标审计                       PERSIST: understanding.md §9 + spec entries
+</states>
+
+<transitions>
+S_INTAKE → S_INTAKE       : -c + session found → A_RESUME_SESSION
+S_INTAKE → S_ARCHAEOLOGY  : issue parsed → A_INTAKE
+S_INTAKE → S_INTAKE       : no issue, no session → AskUserQuestion
+
+S_ARCHAEOLOGY → S_EXPLORE     : A_ARCHAEOLOGY complete
+S_EXPLORE     → S_DIAGNOSE    : A_EXPLORE complete
+
+S_DIAGNOSE → S_FIX          : root cause confirmed, !skip_fix
+S_DIAGNOSE → S_GENERALIZE   : root cause confirmed, skip_fix, !skip_generalize
+S_DIAGNOSE → S_RECORD       : root cause confirmed, skip_fix, skip_generalize
+S_DIAGNOSE → S_DIAGNOSE     : all hypotheses failed, retries < 3 → A_ESCALATE_DIAGNOSIS
+S_DIAGNOSE → S_RECORD       : all hypotheses failed, retries >= 3 → mark INCONCLUSIVE
+
+S_FIX     → S_CONFIRM       : fix implemented
+S_CONFIRM → S_GENERALIZE    : confirmed, !skip_generalize
+S_CONFIRM → S_RECORD        : confirmed, skip_generalize
+S_CONFIRM → S_FIX           : needs_rework
+
+S_GENERALIZE → S_DISCOVER   : similar code found
+S_GENERALIZE → S_RECORD     : no similar code
+
+S_DISCOVER → S_DIAGNOSE     : discovery finds new bug worth investigating → cross_phase_loops++
+S_DISCOVER → S_FIX          : discovery finds same-pattern bug, fix template applies, !skip_fix → cross_phase_loops++
+S_DISCOVER → S_RECORD       : triage complete AND remaining_actionable == 0
+S_DISCOVER → S_RECORD       : loops >= max_loops → MUST log each unfixed item with specific reason (blanket "pre-existing" is forbidden)
+
+S_RECORD   → END            : A_RECORD complete
+</transitions>
+
+<actions>
+
+### A_INTAKE
+1. Parse arguments: issue description, flags
+2. Generate slug, create `SESSION_DIR`
+3. Search: `maestro search "<keywords>"` + Glob prior sessions + ARCHITECTURE.md + Grep keywords
+4. Derive `phase_goals[]` from flags (apply `skip_when`)
+5. Write `session.json` + `understanding.md` §1
+6. Emit Goal Prompt (see Appendix)
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): INTAKE — 目标解析与上下文加载"`
+
+### A_RESUME_SESSION
+Find latest session via Glob → read `session.json` → display summary → jump to `current_state`.
+
+### A_ARCHAEOLOGY
+**Git archaeology (2 parallel Agents):**
+
+| Agent | Task |
+|-------|------|
+| Timeline | `git log --oneline -20 -- {files}` → change timeline |
+| Blame | Top 3 suspicious files: `git blame -L {region}` → critical paths |
+
+Append findings to `evidence.ndjson` (phase: "archaeology").
+
+**CLI-assisted change review:**
+```bash
+maestro delegate "PURPOSE: Review recent modifications related to: {issue}
+TASK: Analyze intent behind changes | Identify risky modifications | Flag potential bug sources
+MODE: analysis
+CONTEXT: @{relevant_files} | Git log: {top_10_commits}
+EXPECTED: JSON [{commit_sha, risk_level, analysis, could_cause_issue, explanation}]
+CONSTRAINTS: Focus on behavioral changes, not formatting
+" --role analyze --mode analysis
+```
+Execute with `run_in_background: true`, then wait for callback (do NOT halt the Odyssey flow). Append results to evidence.
+
+Update `understanding.md` §2.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): ARCHAEOLOGY — git 考古分析"`
+
+### A_EXPLORE
+Skip if no enabled CLI tools (W006).
+
+```bash
+maestro delegate "PURPOSE: Gather codebase evidence for: {issue}
+TASK: Trace call chains | Find recent changes | Identify error gaps | Check similar patterns
+MODE: analysis
+CONTEXT: @**/*
+EXPECTED: JSON {call_chains, recent_changes, error_gaps, similar_patterns}
+CONSTRAINTS: Max 20 entries/category | Symptom-related code paths
+Symptoms: {issue}  Archaeology hints: {suspicious_commits}
+" --role explore --mode analysis
+```
+Execute with `run_in_background: true`, then wait for callback (do NOT halt the Odyssey flow).
+
+Parse → write `explore.json` + append `evidence.ndjson` (phase: "explore"). Update §3. Mark G2 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): EXPLORE — 代码探索完成"`
+
+### A_DIAGNOSE
+1. **Form hypotheses** from evidence (archaeology + explore), ranked [HIGH]/[MEDIUM]/[LOW] → §4
+2. **Test each** (rank order): design test → execute → append evidence (phase: "diagnosis")
+3. **Decision journal**: ambiguity → evidence (phase: "decision"); Normal: AskUserQuestion | `-y`: defer
+4. **Root cause**: confirmed → `session.json.root_cause` + §5. Mark G1 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): DIAGNOSE — 根因确认"`
+
+### A_ESCALATE_DIAGNOSIS
+Increment `diagnosis_retries`. If < 3: broaden scope via `maestro delegate --role analyze` (same delegate format), form new hypotheses, return to S_DIAGNOSE. If >= 3: Normal → AskUserQuestion (broaden/new/INCONCLUSIVE) | `-y` → auto INCONCLUSIVE, proceed to S_RECORD. See Appendix: `-y` behavior.
+
+### A_FIX
+1. Present root cause + proposed fix. Normal: AskUserQuestion | `-y`: auto proceed (see Appendix)
+2. Implement fix
+3. Record in evidence (phase: "decision")
+
+📌 **Auto-commit**: `git add -A && git commit -m "odyssey-debug({slug}): FIX — {修复摘要}"`
+
+### A_CONFIRM
+1. **Tests**: auto-detect framework, run covering tests
+2. **CLI fix review**:
+```bash
+maestro delegate "PURPOSE: Review fix for: {issue}
+TASK: Verify correctness | Check regressions | Assess completeness | Review edge cases
+MODE: analysis
+CONTEXT: @{modified_files} | Root cause: {summary} | Diff: {git_diff}
+EXPECTED: JSON {verdict, findings [{severity, description, suggestion}], regression_risk}
+CONSTRAINTS: Focus on correctness, not style
+" --role review --mode analysis
+```
+Execute with `run_in_background: true`, then wait for callback (do NOT halt the Odyssey flow).
+
+3. Write `session.json.confirmation`: `{test_result, cli_review, overall: "confirmed|needs_rework", timestamp}`
+4. Update §6. `needs_rework` → S_FIX. `confirmed` → mark G3 done, advance.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): CONFIRM — 修复验证"`
+
+### A_GENERALIZE
+举一反三: multi-layer pattern extraction → 4-agent scan → cross-layer dedup → iterative deepening.
+
+**Pattern extraction** from root cause + fix:
+
+| Layer | Method | Example |
+|-------|--------|---------|
+| Syntax | Regex patterns (direct Grep) | `eval(`, missing `await`, unclosed resource |
+| Semantic | Anti-pattern description (Agent-driven) | Unhandled async errors, unvalidated input |
+| Structural | Architecture-level (file/module similarity) | Same import structure, missing override |
+
+Write `session.json.patterns[]`: `[{id, source, layer, signature, description, risk, fix_template}]`
+
+**4-agent parallel codebase scan:**
+
+| Agent | Strategy | Input | Scope |
+|-------|----------|-------|-------|
+| Syntax grep | Grep syntax-layer regex | P*.signature | Full project |
+| Semantic scan | Understand + check anti-pattern | P*.description | Related modules |
+| Structural match | Find structurally similar files | Buggy file structure | Full project |
+| Historical grep | `git log -S "{pattern}"` | P*.signature | Full git history |
+
+Returns: `[{pattern_id, file, line, context, risk_level, layer, confidence}]`
+
+**Cross-layer dedup**: same file:line multi-layer hit → boost confidence | single-layer → `needs_review` | historical hit on fixed record → `regression_risk`
+
+**Iterative deepening**: module with ≥3 hits → targeted deep scan (max 1 round).
+
+**Quality Gate** (self-iteration) → if insufficient, expand and re-scan.
+
+Write §7 + `session.json.generalization_stats`: `{patterns_extracted, total_hits, cross_layer_confirmed, regression_risks, by_layer, deepening_triggered}`. Mark G4 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): GENERALIZE — 泛化扫描完成"`
+
+### A_DISCOVER
+1. **Triage** each hit: read ±10 lines context → classify `safe`/`risk`/`bug`
+2. **Route**:
+   - `bug` + directly fixable → **fix immediately** (not just log an issue) → back to S_FIX
+   - `bug` + requires cross-module/architectural decision → create issue (with fix suggestion + impact analysis)
+   - `risk` → evaluate if guard/validation can mitigate directly; if yes, fix it
+   - `safe` → mark skip
+   See Appendix `-y` behavior table. Append evidence (phase: "discovery" + "decision")
+3. Update §8. Mark G5 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): DISCOVER — 发现分类完成"`
+
+### A_RECORD
+1. Finalize `understanding.md` §9
+2. **Write learnings** to understanding.md §9: 按 Knowledge Persistence 表分类记录（临时），completion summary 列出建议的 `/spec-add` 命令
+3. Mark G6 done. Process pending decisions: Normal → AskUserQuestion | `-y` → skip (show deferred count)
+4. **Goal audit**: all `completion_confirmed` true → `phase_goals_all_done = true`. Any false: Normal → AskUserQuestion (回退/跳过/接受) | `-y` → auto accept
+5. **Completion**: `current_state = "COMPLETED"`, emit summary:
+```
+--- DEBUG ODYSSEY COMPLETE ---
+Issue:      {issue}
+Root cause: {root_cause.hypothesis}
+Fix:        {applied|skipped|inconclusive}
+Patterns:   {patterns_extracted} ({by_layer} distribution)
+Scan hits:  {total_hits} ({cross_layer_confirmed} cross-layer confirmed)
+Issues:     {N} created
+Decisions:  {N} resolved, {M} pending, {K} deferred
+Learnings:  {N} spec entries persisted
+Self-iter:  {N} quality gate rounds across {M} stages
+Goals:      {done}/{total} confirmed ({skipped} skipped)
+---
+```
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): RECORD — 会话总结与知识沉淀"`
+
+</actions>
+
+<appendix>
+
+### Goal Prompt Template
+
+**时机守卫：仅在 A_INTAKE 完成后显示一次。A_RECORD 完成时禁止重新显示。**
+
+```
+📋 Debug Odyssey 会话已创建。可随时复制以下 /goal 设定终止条件：
+
+/goal 完成以下目标：
+{for each G in phase_goals where status != "skipped":}
+- {G.id}: {G.goal} — 完成条件: {G.done_when}
+{end for}
+穷尽迭代：直到根因确认（或明确 INCONCLUSIVE）且修复验证通过
+且泛化扫描穷尽（所有 hits 已分类处理）且 phase_goals_all_done=true 才停。
+假设失败时扩范围→换视角→升级工具，不轻易放弃。
+泛化发现的同类 bug 全部修复或创建 issue，不允许遗留。
+遇到 phase=decision 的 pending 必须 AskUserQuestion，不得自行 resolve。
+```
+
+Odyssey 输出提示词后继续执行不阻塞。`/goal` 由用户任意时刻输入。
+
+### `-y` Auto-Confirm Behavior
+
+| Decision Point | Normal | `-y` mode |
+|---------------|--------|-----------|
+| A_DIAGNOSE ambiguity | AskUserQuestion blocks | record `deferred`, best-effort continue |
+| A_ESCALATE 3-strike | AskUserQuestion 3-way | auto INCONCLUSIVE |
+| A_FIX direction | AskUserQuestion confirm | auto proceed with suggested fix |
+| A_DISCOVER bug triage | AskUserQuestion route | auto create issue |
+| A_DISCOVER ambiguous | AskUserQuestion batch | all `deferred` |
+| A_RECORD decisions | AskUserQuestion per-item | skip, show deferred count |
+| A_RECORD goal audit | AskUserQuestion 3-way | auto accept current state |
+
+`deferred` items shown as "待决策" in completion summary; recoverable via `-c`.
+
+### Phase Goal Lifecycle
+
+`pending → done (confirmed=true)` normal | `pending → skipped (confirmed=true)` flags/manual | `pending → failed (confirmed=false)` INCONCLUSIVE
+
+`phase_goals_all_done = true` only when ALL goals have `completion_confirmed == true`.
+
+</appendix>
+
+</state_machine>
+
+<error_codes>
+| Code | Severity | Condition | Recovery |
+|------|----------|-----------|----------|
+| E001 | error | No issue and no session to resume | Provide issue or use -c |
+| E003 | error | Resume but no session found | Start new session |
+| E004 | error | Delegate execution failed | Retry or proceed without CLI |
+| W001 | warning | No relevant git history | Proceed with limited context |
+| W002 | warning | All hypotheses inconclusive after 3 retries | INCONCLUSIVE |
+| W003 | warning | Generalization scan 0 hits | Skip discovery |
+| W004 | warning | Delegate parse failed | Use raw output |
+| W005 | warning | Pending decisions unresolved | Filter evidence.ndjson phase=decision |
+| W006 | warning | CLI exploration skipped (no tools) | Proceed without explore.json |
+</error_codes>
+
+<success_criteria>
+- [ ] Session directory created with 4 output files
+- [ ] Prior knowledge searched (maestro search + sessions + architecture)
+- [ ] Git archaeology (log + blame) + CLI change review, evidence.ndjson phase=archaeology
+- [ ] CLI exploration, explore.json written, evidence phase=explore
+- [ ] Hypotheses formed from archaeology + explore, tested and logged phase=diagnosis
+- [ ] Root cause declared with evidence refs
+- [ ] understanding.md tracks all 9 sections progressively
+- [ ] Fix implemented + confirmed with test + CLI review (unless --skip-fix)
+- [ ] Multi-layer patterns (syntax/semantic/structural) extracted (unless --skip-generalize)
+- [ ] 4-agent scan + cross-layer dedup + iterative deepening for ≥3 hits/module
+- [ ] Discoveries classified and routed (fix/issue/decision/skip)
+- [ ] **Every unfixed finding has individual classification and reason** — blanket "pre-existing" labels are forbidden
+- [ ] Decision journal: all human-judgment items in evidence.ndjson phase=decision
+- [ ] phase_goals derived from flags, skip_when applied, each phase marks its goal
+- [ ] Goal audit in A_RECORD — unmet goals surfaced, phase_goals_all_done set correctly
+- [ ] Goal Prompt displayed once after session creation
+- [ ] `-y`: all decisions auto-resolve/defer, deferred count in summary
+- [ ] State saved at each transition (resumable via -c)
+- [ ] Quality Gate self-iteration when insufficient, logged in self_iteration_log
+- [ ] Spec entries persisted for reusable learnings
+- [ ] Completion summary with all stats
+</success_criteria>
+
+<next_step_routing>
+| Condition | Next step |
+|-----------|-----------|
+| Issues from discoveries | `/manage-issue list --source debug-odyssey` |
+| Pattern worth documenting | `/learn-decompose <module>` |
+| Fix needs formal review | `/quality-review <phase>` |
+| Second opinion on root cause | `/learn-second-opinion <understanding.md>` |
+| Related question | `/learn-investigate "<question>"` |
+| Decisions still pending | Filter evidence.ndjson phase=decision status=pending |
+</next_step_routing>