maestro-flow 0.5.44 → 0.5.46

package/.agy/skills/odyssey-improve/SKILL.md

@@ -14,284 +14,370 @@ allowed-tools:
   - view_file
   - write_to_file
 ---
-<base>@~/.maestro/workflows/odyssey-base.md</base>
-
-<purpose>
-survey → 6-dimension audit → diagnose → fix → verify → generalize → discover → persist.
-Exhaustive iteration until zero remaining actionable findings.
-</purpose>
-
-<boundary>
-**In scope:** Runtime quality improvement — performance/security/architecture/reliability/observability/maintainability audit → diagnose → fix → generalize.
-**Out of scope:** UI visual → `/odyssey-ui` | New features → `/odyssey-planex` | Single bug → `/odyssey-debug` | Style review → `/odyssey-review-test-fix`
-**Exploration freedom:** Free exploration within boundary — profiling, security scanning, architecture analysis, dependency audit.
-**Zero-residual:** 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.
-</boundary>
-
-<context>
-$ARGUMENTS
-
-**Target resolution:**
-| Input | Resolution |
-|-------|-----------|
-| Module/dir path | Audit that module |
-| `HEAD` / `staged` | Review changes in diff |
-| Feature area keyword | Resolve to related files |
-| `--all` | Full project scan (use with caution) |
-
-**Flags:** `--dimensions <list>` dimension subset | `--fix-threshold <severity>` fix cutoff (default: all) | `--skip-fix` audit+diagnose only | `--skip-generalize` skip S_GENERALIZE+S_DISCOVER | `--auto` no delegate confirmation | `-y` auto-confirm | `-c` resume | `--heartbeat` /loop heartbeat
-
-**Dimensions (6):**
-1. **performance** — hot paths, N+1 queries, memory allocation, cache efficiency, bundle size, lazy loading
-2. **security** — OWASP Top 10, injection, auth bypass, data exposure, dependency vulnerabilities, secrets
-3. **architecture** — layer violations, circular dependencies, coupling metrics, interface contracts, SRP violations
-4. **reliability** — error handling gaps, retry logic, timeout handling, graceful degradation, resource cleanup
-5. **observability** — logging coverage, metric gaps, trace propagation, error reporting, health checks
-6. **maintainability** — code complexity (cyclomatic), dead code, test coverage gaps, documentation debt
-
-**Session**: `.workflow/scratch/{YYYYMMDD}-improve-odyssey-{slug}/`
-**Output**: `session.json` | `evidence.ndjson` | `understanding.md`
-
-**session.json — improve-specific fields:**
-```json
-{ "target": "", "dimensions": [], "baseline_metrics": {},
-  "audit_result": {}, "diagnoses": [], "confirmation": null,
-  "generalization_stats": null }
-```
-
-**evidence.ndjson phases:** `survey|audit|diagnosis|fix|discovery|decision|self-iteration`
-- `survey`: `category` (dependency|complexity|coverage|error_pattern), `detail`
-- `audit`: `dimension`, `severity`, `measurement`
-- `diagnosis`: `finding_ref`, `hypothesis`, `result` (confirmed|disproved|inconclusive), `root_cause`
-- `fix`: `finding_ref`, `change_summary`, `risk`
-- `discovery`: `file`, `line`, `classification` (safe|risk|issue), `action` (fix|issue|decision|skip)
-- `decision`: `question`, `options`, `context`, `status` (pending|resolved|deferred), `resolution`
-- `self-iteration`: `stage`, `round`, `assessment`, `expansion`
-
-**phase_goals[]:**
-
-| ID | Goal | Phase | skip_when |
-|----|------|-------|-----------|
-| G1 | Survey completed | S_SURVEY | — |
-| G2 | Audit completed | S_AUDIT | — |
-| G3 | Diagnosis completed | S_DIAGNOSE | — |
-| G4 | Zero remaining: all findings fixed and verified | S_VERIFY | skip_fix |
-| G5 | Pattern generalized | S_GENERALIZE | skip_generalize |
-| G6 | Discoveries triaged | S_DISCOVER | skip_generalize |
-| G7 | Learnings persisted | S_RECORD | — |
-
-**understanding.md — 9 sections:**
-1. Target & Baseline ← S_INTAKE | 2. Current State Survey ← S_SURVEY | 3. Audit Findings ← S_AUDIT
-4. Root Cause Diagnosis ← S_DIAGNOSE | 5. Fix & Verification ← S_FIX+S_VERIFY
-6. Generalization ← S_GENERALIZE | 7. Discoveries ← S_DISCOVER
-8. Improvement Metrics ← S_RECORD (before/after) | 9. Engineering Learnings ← S_RECORD
-
-**Knowledge Persistence categories (§9):**
-
-| Category | Content | Follow-up |
-|----------|---------|-----------|
-| Performance pattern | Bottleneck type + fix approach + measurement method | `/spec-add coding` |
-| Security rule | Vulnerability class + fix + prevention method | `/spec-add debug` |
-| Architecture constraint | Violation description + correct boundary + check method | `/spec-add arch` |
-| Reliability pattern | Failure mode + handling strategy + verification method | `/spec-add coding` |
-</context>
-
-<invariants>
-All invariants (evidence append-only, session-as-state, phase goal tracking, auto-commit, zero-residual) defined in base.
-</invariants>
-
-<self_iteration>
-Applies to: **S_SURVEY, S_AUDIT, S_DIAGNOSE, S_GENERALIZE**. Logic in base.
-</self_iteration>
-
-<state_machine>
-
-<states>
-S_INTAKE → S_SURVEY → S_AUDIT → S_DIAGNOSE → S_FIX → S_VERIFY → S_GENERALIZE → S_DISCOVER → S_RECORD → END
-</states>
-
-<transitions>
-S_INTAKE → S_INTAKE      : -c + session found → A_RESUME
-S_INTAKE → S_SURVEY      : target resolved → A_INTAKE
-S_INTAKE → S_INTAKE      : no target → ask_question
-
-S_SURVEY   → S_AUDIT       : complete
-
-S_AUDIT → S_DIAGNOSE     : critical/high findings exist
-S_AUDIT → S_GENERALIZE   : no critical/high, !skip_generalize
-S_AUDIT → S_RECORD       : no findings OR skip_generalize
-
-S_DIAGNOSE → S_FIX          : root causes identified, !skip_fix
-S_DIAGNOSE → S_GENERALIZE   : root causes identified, skip_fix, !skip_generalize
-S_DIAGNOSE → S_RECORD       : root causes identified, skip_fix, skip_generalize
-S_DIAGNOSE → S_DIAGNOSE     : hypotheses failed, retries < 3 → A_ESCALATE_DIAGNOSIS
-S_DIAGNOSE → S_RECORD       : retries >= 3 → INCONCLUSIVE
-
-S_FIX      → S_VERIFY      : fix implemented
-
-S_VERIFY → S_GENERALIZE   : verified, !skip_generalize
-S_VERIFY → S_RECORD       : verified, skip_generalize
-S_VERIFY → S_FIX          : needs_rework
-
-S_GENERALIZE → S_DISCOVER   : hits found
-S_GENERALIZE → S_RECORD     : no hits
-
-S_DISCOVER → S_DIAGNOSE     : new critical issue → cross_phase_loops++
-S_DISCOVER → S_FIX          : same-pattern fix, !skip_fix → cross_phase_loops++
-S_DISCOVER → S_RECORD       : remaining_actionable == 0
-S_DISCOVER → S_RECORD       : loops >= max_loops → MUST log per-item reasons
-
-S_RECORD   → END            : complete
-</transitions>
-
-<actions>
-
-### A_INTAKE
-1. Parse arguments: target description, flags, `--dimensions` subset
-2. Generate slug, create `SESSION_DIR`
-3. Search: `maestro search "<keywords>"` + Glob prior sessions + ARCHITECTURE.md + spec load coding/debug
-4. **Baseline capture**: Record current metrics (test pass rate, bundle size, dependency count, complexity hotspots) to `session.json.baseline_metrics`
-5. Derive `phase_goals[]` from flags
-6. Write `session.json` + `understanding.md` §1, emit Goal Prompt
-
-Commit: `"odyssey-improve({slug}): INTAKE — parse target and capture baseline"`
-
-### A_RESUME
-Glob latest session → read `session.json` → display summary → jump to `current_state`.
-
-### A_SURVEY
-Current state survey — understand what exists before proposing changes.
-
-1. Dependency audit (package.json/lock), complexity scan (size/nesting), test coverage map, error handling scan (empty catch, unhandled promise)
-2. **CLI-assisted** (optional): `maestro delegate --role analyze --mode analysis` for dependency health, complexity hotspots, coverage gaps, error patterns. Execute `run_in_background: true`.
-3. Evidence phase=survey. Update §2. Mark G1.
-
-Commit: `"odyssey-improve({slug}): SURVEY — current state analysis"`
-
-### A_AUDIT
-Spawn 6 parallel Agents (one per dimension, or `--dimensions` subset).
-Each returns: `[{title, severity, dimension, file, line, description, suggestion, measurement}]`
-
-Merge → evidence phase=audit. Write `session.json.audit_result`.
-Update §3 (findings by dimension + severity matrix). Mark G2.
-
-Commit: `"odyssey-improve({slug}): AUDIT — multi-dimension review"`
-
-### A_DIAGNOSE
-Root cause analysis for critical/high findings — don't fix symptoms.
-
-1. Group by dimension, prioritize by severity. For each: hypothesis → trace code path + git history → evidence phase=diagnosis
-2. Ambiguity → evidence phase=decision; Normal: ask_question | `-y`: defer
-3. CLI-assisted for complex findings: `maestro delegate --role analyze --mode analysis` (`run_in_background: true`)
-4. Write `session.json.diagnoses[]`. Update §4. Mark G3.
-
-Commit: `"odyssey-improve({slug}): DIAGNOSE — root cause analysis"`
-
-### A_ESCALATE_DIAGNOSIS
-`retries++`. < 3: `maestro delegate --role analyze`, new hypotheses, → S_DIAGNOSE. >= 3: Normal → ask_question | `-y` → INCONCLUSIVE → S_RECORD.
-
-### A_FIX
-1. Exhaustive fix: ALL diagnosed issues by severity tier (critical → high → medium → low within fix_threshold), one dimension at a time. After each tier, re-verify **current tier's dimension only** (not all dimensions) — new findings at same or higher severity append to current tier. Cross-dimension regression checks run once at S_VERIFY after all tiers complete.
-2. For each fix: implement → evidence phase=fix
-3. Normal: ask_question per-fix confirmation | `-y`: auto-proceed, record `deferred`
-
-Commit: `"odyssey-improve({slug}): FIX — improvements applied"`
-
-### A_VERIFY
-1. Run tests covering modified areas
-2. Re-capture metrics, compare with `session.json.baseline_metrics`
-3. CLI-assisted: `maestro delegate --role review --mode analysis` (`run_in_background: true`)
-4. `needs_rework` → S_FIX. `verified` → mark G4, advance.
-5. Write `session.json.confirmation`. Update §5 (before/after metrics table).
-
-Commit: `"odyssey-improve({slug}): VERIFY — improvements verified"`
-
-### A_GENERALIZE, A_DISCOVER, A_RECORD
-Base shared_actions. Improve overrides:
-- **A_GENERALIZE** pattern source: diagnoses + fixes
-- **A_RECORD** §8: improvement metrics (before/after comparison from baseline_metrics vs current). §9: learnings per Knowledge Persistence table. Completion summary lists suggested `/spec-add` commands.
-
-**Completion summary:**
-```
---- IMPROVE ODYSSEY COMPLETE ---
-Target:      {target}
-Dimensions:  {dimensions}
-Findings:    {critical}C / {high}H / {medium}M / {low}L
-Diagnosed:   {count}
-Fixed:       {count} ({verified} verified)
-Metrics:     {improved} improved / {regressed} regressed
-Patterns:    {count} ({by_layer})
-Scan hits:   {total} ({cross_layer_confirmed} confirmed)
-Issues:      {N} created
-Decisions:   {N} resolved, {M} pending, {K} deferred
-Learnings:   {N} persisted
-Self-iter:   {N} rounds across {M} stages
-Cross-loops: {N}
-Goals:       {done}/{total} ({skipped} skipped)
----
-```
-
-</actions>
-
-<appendix>
-
-### `-y` improve-specific points
-
-| Decision Point | Normal | `-y` |
-|---------------|--------|------|
-| A_FIX improvement confirmation | ask_question | auto-proceed, `deferred` |
-| A_DIAGNOSE ambiguity | ask_question | best-effort, `deferred` |
-| A_ESCALATE 3-strike | ask_question 3-way | auto INCONCLUSIVE |
-| A_DISCOVER hit routing | ask_question | auto create issue |
-| A_DISCOVER ambiguous items | ask_question | all `deferred` |
-
-`deferred` items shown in completion summary; recoverable via `-c`.
-
-### Goal Prompt convergence rules
-
-```
-Exhaust iteration until all findings actioned (fix/issue/decision)
-and phase_goals_all_done=true.
-Fix by severity tiers, re-verify after each tier.
-Baseline captured before fix, compared after to confirm improvement.
-Pending decisions must ask_question — no silent resolve.
-```
-
-</appendix>
-
-</state_machine>
-
-<error_codes>
-| Code | Severity | Condition | Recovery |
-|------|----------|-----------|----------|
-| E001 | error | No target specified | Provide target or use -c |
-| E002 | error | Target path not found | Check path |
-| W001 | warning | No dependency manifest found | Proceed without dep audit |
-| W002 | warning | Some dimension agents failed | Partial audit coverage |
-</error_codes>
-
-<success_criteria>
-- [ ] Target resolved, baseline metrics captured
-- [ ] Survey + 6-dimension audit with structured findings and severity matrix
-- [ ] Root causes diagnosed for critical/high findings
-- [ ] Improvements implemented and verified with before/after metrics (unless --skip-fix)
-- [ ] Multi-layer generalization + cross-phase loops (unless --skip-generalize)
-- [ ] Every unfixed finding has individual classification and reason
-- [ ] understanding.md §8 (metrics) and §9 (learnings) completed
-- [ ] phase_goals G1-G7 tracked and audited
-- [ ] Session resumable via -c
-- [ ] Completion summary
-</success_criteria>
-
-<next_step_routing>
-| Condition | Next |
-|-----------|------|
-| Security findings need deep investigation | `/odyssey-debug "<finding>"` |
-| UI-related findings | `/odyssey-ui "<component>"` |
-| Issues created from discoveries | `/manage-issue list --source improve-odyssey` |
-| Architecture pattern to document | `/spec-add arch "..."` |
-| Performance pattern to persist | `/spec-add coding "..."` |
-| Formal review of changes | `/odyssey-review-test-fix <changed-files>` |
-| Pending decisions | Filter evidence phase=decision status=pending |
-</next_step_routing>
-</output>
+<base>@~/.maestro/workflows/odyssey-base.md</base>
+
+<purpose>
+survey → 6-dimension audit → diagnose → fix → verify → generalize → discover → persist.
+Exhaustive iteration until zero remaining actionable findings.
+</purpose>
+
+<boundary>
+**In scope:** Runtime quality improvement — performance/security/architecture/reliability/observability/maintainability audit → diagnose → fix → generalize.
+**Out of scope:** UI visual → `/odyssey-ui` | New features → `/odyssey-planex` | Single bug → `/odyssey-debug` | Style review → `/odyssey-review-test-fix`
+**Exploration freedom:** Free exploration within boundary — profiling, security scanning, architecture analysis, dependency audit.
+**Zero-residual:** 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.
+</boundary>
+
+<context>
+$ARGUMENTS
+
+**Target resolution:**
+| Input | Resolution |
+|-------|-----------|
+| Module/dir path | Audit that module |
+| `HEAD` / `staged` | Review changes in diff |
+| Feature area keyword | Resolve to related files |
+| `--all` | Full project scan (use with caution) |
+
+**Flags:** `--dimensions <list>` dimension subset | `--fix-threshold <severity>` fix cutoff (default: all) | `--skip-fix` audit+diagnose only | `--skip-generalize` skip S_GENERALIZE+S_DISCOVER | `--auto` no delegate confirmation | `-y` auto-confirm | `-c` resume | `--heartbeat` /loop heartbeat
+
+**Dimensions (6):**
+1. **performance** — hot paths, N+1 queries, memory allocation, cache efficiency, bundle size, lazy loading
+2. **security** — OWASP Top 10, injection, auth bypass, data exposure, dependency vulnerabilities, secrets
+3. **architecture** — layer violations, circular dependencies, coupling metrics, interface contracts, SRP violations
+4. **reliability** — error handling gaps, retry logic, timeout handling, graceful degradation, resource cleanup
+5. **observability** — logging coverage, metric gaps, trace propagation, error reporting, health checks
+6. **maintainability** — code complexity (cyclomatic), dead code, test coverage gaps, documentation debt
+
+**Session**: `.workflow/scratch/{YYYYMMDD}-improve-odyssey-{slug}/`
+**Output**: `session.json` | `evidence.ndjson` | `understanding.md`
+
+**session.json — improve-specific fields:**
+```json
+{ "target": "", "dimensions": [], "baseline_metrics": {},
+  "audit_result": {}, "diagnoses": [], "confirmation": null,
+  "generalization_stats": null }
+```
+
+**evidence.ndjson phases:** `survey|audit|diagnosis|fix|discovery|decision|self-iteration`
+- `survey`: `category` (dependency|complexity|coverage|error_pattern), `detail`
+- `audit`: `dimension`, `severity`, `measurement`
+- `diagnosis`: `finding_ref`, `hypothesis`, `result` (confirmed|disproved|inconclusive), `root_cause`
+- `fix`: `finding_ref`, `change_summary`, `risk`
+- `discovery`: `file`, `line`, `classification` (safe|risk|issue), `action` (fix|issue|decision|skip)
+- `decision`: `question`, `options`, `context`, `status` (pending|resolved|deferred), `resolution`
+- `self-iteration`: `stage`, `round`, `assessment`, `expansion`
+
+**phase_goals[]:**
+
+| ID | Goal | Phase | skip_when |
+|----|------|-------|-----------|
+| G1 | Survey completed | S_SURVEY | — |
+| G2 | Audit completed | S_AUDIT | — |
+| G3 | Diagnosis completed | S_DIAGNOSE | — |
+| G4 | Zero remaining: all findings fixed and verified | S_VERIFY | skip_fix |
+| G5 | Pattern generalized | S_GENERALIZE | skip_generalize |
+| G6 | Discoveries triaged | S_DISCOVER | skip_generalize |
+| G7 | Learnings persisted | S_RECORD | — |
+
+**understanding.md — 9 sections:**
+1. Target & Baseline ← S_INTAKE | 2. Current State Survey ← S_SURVEY | 3. Audit Findings ← S_AUDIT
+4. Root Cause Diagnosis ← S_DIAGNOSE | 5. Fix & Verification ← S_FIX+S_VERIFY
+6. Generalization ← S_GENERALIZE | 7. Discoveries ← S_DISCOVER
+8. Improvement Metrics ← S_RECORD (before/after) | 9. Engineering Learnings ← S_RECORD
+
+**Knowledge Persistence categories (§9):**
+
+| Category | Content | Follow-up |
+|----------|---------|-----------|
+| Performance pattern | Bottleneck type + fix approach + measurement method | `/spec-add coding` |
+| Security rule | Vulnerability class + fix + prevention method | `/spec-add debug` |
+| Architecture constraint | Violation description + correct boundary + check method | `/spec-add arch` |
+| Reliability pattern | Failure mode + handling strategy + verification method | `/spec-add coding` |
+</context>
+
+<invariants>
+All invariants (evidence append-only, session-as-state, phase goal tracking, auto-commit, zero-residual) defined in base.
+6. **Generalize is mandatory** — S_GENERALIZE and S_DISCOVER execute unless `skip_generalize == true`. "All verified" or context pressure are NOT valid reasons to skip. The phase itself determines whether patterns exist.
+</invariants>
+
+<self_iteration>
+Applies to: **S_SURVEY, S_AUDIT, S_DIAGNOSE, S_GENERALIZE**. Logic in base.
+</self_iteration>
+
+<state_machine>
+
+<states>
+S_INTAKE → S_SURVEY → S_AUDIT → S_DIAGNOSE → S_FIX → S_VERIFY → S_GENERALIZE → S_DISCOVER → S_RECORD → END
+</states>
+
+<transitions>
+S_INTAKE → S_INTAKE      : -c + session found → A_RESUME
+S_INTAKE → S_SURVEY      : target resolved → A_INTAKE
+S_INTAKE → S_INTAKE      : no target → ask_question
+
+S_SURVEY   → S_AUDIT       : complete
+
+S_AUDIT → S_DIAGNOSE     : critical/high findings exist
+S_AUDIT → S_GENERALIZE   : no critical/high, !skip_generalize
+S_AUDIT → S_RECORD       : no findings OR skip_generalize
+
+S_DIAGNOSE → S_FIX          : root causes identified, !skip_fix
+S_DIAGNOSE → S_GENERALIZE   : root causes identified, skip_fix, !skip_generalize
+S_DIAGNOSE → S_RECORD       : root causes identified, skip_fix, skip_generalize
+S_DIAGNOSE → S_DIAGNOSE     : hypotheses failed, retries < 3 → A_ESCALATE_DIAGNOSIS
+S_DIAGNOSE → S_RECORD       : retries >= 3 → INCONCLUSIVE
+
+S_FIX      → S_VERIFY      : fix implemented
+
+S_VERIFY → S_GENERALIZE   : verified, !skip_generalize
+S_VERIFY → S_RECORD       : verified, skip_generalize
+S_VERIFY → S_FIX          : needs_rework
+
+S_GENERALIZE → S_DISCOVER   : hits found
+S_GENERALIZE → S_RECORD     : all 3 layers scanned with evidence, total_hits == 0
+
+S_DISCOVER → S_DIAGNOSE     : new critical issue → cross_phase_loops++
+S_DISCOVER → S_FIX          : same-pattern fix, !skip_fix → cross_phase_loops++
+S_DISCOVER → S_RECORD       : remaining_actionable == 0
+S_DISCOVER → S_RECORD       : loops >= max_loops → MUST log per-item reasons
+
+S_RECORD   → END            : complete
+</transitions>
+
+<actions>
+
+### A_INTAKE
+1. Parse arguments: target description, flags, `--dimensions` subset
+2. Generate slug, create `SESSION_DIR`
+3. Search: `maestro search "<keywords>"` + Glob prior sessions + ARCHITECTURE.md + spec load coding/debug
+4. **Baseline capture**: Record current metrics (test pass rate, bundle size, dependency count, complexity hotspots) to `session.json.baseline_metrics`
+5. Derive `phase_goals[]` from flags
+6. Write `session.json` + `understanding.md` §1, emit Goal Prompt
+
+Commit: `"odyssey-improve({slug}): INTAKE — parse target and capture baseline"`
+
+### A_RESUME
+Glob latest session → read `session.json` → display summary → jump to `current_state`.
+
+### A_SURVEY
+Current state survey — understand what exists before proposing changes.
+
+1. Dependency audit (package.json/lock), complexity scan (size/nesting), test coverage map, error handling scan (empty catch, unhandled promise)
+2. **CLI-assisted** (optional): `maestro delegate --role analyze --mode analysis` for dependency health, complexity hotspots, coverage gaps, error patterns. Execute `run_in_background: true`.
+3. Evidence phase=survey. Update §2. Mark G1.
+
+Commit: `"odyssey-improve({slug}): SURVEY — current state analysis"`
+
+### A_AUDIT
+Spawn 6 parallel Agents (one per dimension, or `--dimensions` subset).
+Each returns: `[{title, severity, dimension, file, line, description, suggestion, measurement}]`
+
+Merge → evidence phase=audit. Write `session.json.audit_result`.
+Update §3 (findings by dimension + severity matrix). Mark G2.
+
+Commit: `"odyssey-improve({slug}): AUDIT — multi-dimension review"`
+
+### A_DIAGNOSE
+Root cause analysis for critical/high findings — don't fix symptoms.
+
+1. Group by dimension, prioritize by severity. For each: hypothesis → trace code path + git history → evidence phase=diagnosis
+2. Ambiguity → evidence phase=decision; Normal: ask_question | `-y`: defer
+3. CLI-assisted for complex findings: `maestro delegate --role analyze --mode analysis` (`run_in_background: true`)
+4. Write `session.json.diagnoses[]`. Update §4. Mark G3.
+
+Commit: `"odyssey-improve({slug}): DIAGNOSE — root cause analysis"`
+
+### A_ESCALATE_DIAGNOSIS
+`retries++`. < 3: `maestro delegate --role analyze`, new hypotheses, → S_DIAGNOSE. >= 3: Normal → ask_question | `-y` → INCONCLUSIVE → S_RECORD.
+
+### A_FIX
+1. Exhaustive fix: ALL diagnosed issues by severity tier (critical → high → medium → low within fix_threshold), one dimension at a time. After each tier, re-verify **current tier's dimension only** (not all dimensions) — new findings at same or higher severity append to current tier. Cross-dimension regression checks run once at S_VERIFY after all tiers complete.
+2. For each fix: implement → evidence phase=fix
+3. Normal: ask_question per-fix confirmation | `-y`: auto-proceed, record `deferred`
+
+Commit: `"odyssey-improve({slug}): FIX — improvements applied"`
+
+### A_VERIFY
+1. Run tests covering modified areas
+2. Re-capture metrics, compare with `session.json.baseline_metrics`
+3. CLI-assisted: `maestro delegate --role review --mode analysis` (`run_in_background: true`)
+4. `needs_rework` → S_FIX. `verified` → mark G4, advance.
+5. Write `session.json.confirmation`. Update §5 (before/after metrics table).
+
+Commit: `"odyssey-improve({slug}): VERIFY — improvements verified"`
+
+### A_GENERALIZE
+
+**MANDATORY — executes unless `skip_generalize == true`. Prior-phase convergence, "all verified," or context pressure are NOT valid skip reasons.**
+
+Pattern source: diagnosed root causes + applied fixes across all dimensions.
+
+**Step 1 — 3-layer pattern extraction** from diagnoses and fixes:
+
+| Layer | Method | Targets |
+|-------|--------|---------|
+| Syntax | Build regex from fix diffs → Grep | N+1 query patterns, missing error type check, unescaped interpolation, inline eval |
+| Semantic | Understand anti-pattern per dimension → Agent scan | Cache bypass in hot path, auth check gap, circular import chain, missing timeout |
+| Structural | Find modules with same shape / dependency graph | Controllers sharing validation shape, middleware with same hook pattern, parallel services |
+
+Write `session.json.patterns[]`: `[{id, source, layer, signature, description, risk, fix_template, confidence}]`
+
+**Thoroughness floor:** ALL 3 layers must be attempted and logged. Each layer records search method, scope, hit count in evidence phase=generalization. "No hits" requires all 3 layers to return 0 with logged evidence.
+
+**Step 2 — 4-agent concurrent scan** (single message, 4 Agents):
+
+| Agent | Strategy | Scope |
+|-------|----------|-------|
+| Syntax grep | Grep regex from pattern signatures | Full project |
+| Semantic scan | Per-dimension anti-pattern check | Related modules |
+| Structural match | Find structurally similar files | Full project |
+| Historical grep | `git log -S` for pattern signatures | Git history |
+
+**Step 3 — Cross-layer dedup:** multi-layer hit → boost | single-layer → `needs_review` | historically fixed → `regression_risk`
+
+**Step 4 — Iterative deepening:** Module with ≥3 hits → targeted deep scan (max 1 round).
+
+**Step 5 — Persist:** Update understanding.md §6 + write `session.json.generalization_stats`:
+```json
+{"patterns_extracted": 0, "total_hits": 0, "cross_layer_confirmed": 0, "regression_risks": 0, "by_layer": {"syntax": 0, "semantic": 0, "structural": 0}, "deepening_triggered": false}
+```
+
+**Transition guard:** `S_GENERALIZE → S_RECORD` requires `by_layer` has entries for all 3 layers with search evidence logged. Mark G5 done.
+
+Commit: `"odyssey-improve({slug}): GENERALIZE — generalization scan complete"`
+
+### A_DISCOVER
+
+**Executes whenever `total_hits > 0`. Cannot be skipped without `skip_generalize == true`.**
+
+1. **Triage** each hit with ±10 lines context → classify:
+   - `bug` — confirmed improvement defect (same dimension)
+   - `risk` — potential degradation or violation
+   - `safe` — false positive (must log individual reason — blanket "pre-existing" forbidden)
+
+2. **Route:**
+
+   | Classification | Action |
+   |---------------|--------|
+   | bug + fix_template applicable | Immediate fix → back to S_FIX |
+   | bug + cross-module decision or no template | Create issue (fix suggestion + impact + dimension) |
+   | risk + guard addable directly | Fix directly |
+   | risk + complex | Create issue |
+   | safe | Skip with logged per-item reason |
+
+   Normal: ask_question per hit | `-y`: auto-fix bugs with fix_template, create issue for rest
+
+3. **Cross-phase loops:** `cross_phase_loops++` on fix/diagnose return. `loops >= max_loops` → must log per-item reasons.
+
+4. Append evidence phase=discovery. Update understanding.md §7. Mark G6 done.
+
+Commit: `"odyssey-improve({slug}): DISCOVER — discovery triage complete"`
+
+### A_RECORD
+
+1. understanding.md §8: improvement metrics — before/after comparison from `baseline_metrics` vs current. Re-capture metrics, build comparison table.
+2. understanding.md §9: learnings by Knowledge Persistence categories:
+   - Performance pattern → `/spec-add coding`
+   - Security rule → `/spec-add debug`
+   - Architecture constraint → `/spec-add arch`
+   - Reliability pattern → `/spec-add coding`
+   Completion summary lists suggested `/spec-add` commands.
+
+3. Mark G7 done. Pending decisions: Normal → ask_question | `-y` → skip (show deferred count).
+
+4. **Goal audit (hardened):**
+   - `done` → confirmed
+   - `skipped` → confirmed ONLY if corresponding `skip_when` flag is true
+   - **Hard rule:** G5 and G6 CANNOT be `skipped` unless `skip_generalize == true`. Pending without flag → `failed` (Normal: ask_question | `-y`: record `failed`)
+   - `phase_goals_all_done = true` only when all goals pass this audit
+
+5. `current_state = "COMPLETED"`, emit completion summary.
+
+**Completion summary:**
+```
+--- IMPROVE ODYSSEY COMPLETE ---
+Target:      {target}
+Dimensions:  {dimensions}
+Findings:    {critical}C / {high}H / {medium}M / {low}L
+Diagnosed:   {count}
+Fixed:       {count} ({verified} verified)
+Metrics:     {improved} improved / {regressed} regressed
+Patterns:    {count} ({by_layer})
+Scan hits:   {total} ({cross_layer_confirmed} confirmed)
+Issues:      {N} created
+Decisions:   {N} resolved, {M} pending, {K} deferred
+Learnings:   {N} persisted
+Self-iter:   {N} rounds across {M} stages
+Cross-loops: {N}
+Goals:       {done}/{total} ({skipped} skipped)
+---
+```
+
+Commit: `"odyssey-improve({slug}): RECORD — summary and knowledge persistence"`
+
+</actions>
+
+<appendix>
+
+### `-y` improve-specific points
+
+| Decision Point | Normal | `-y` |
+|---------------|--------|------|
+| A_FIX improvement confirmation | ask_question | auto-proceed, `deferred` |
+| A_DIAGNOSE ambiguity | ask_question | best-effort, `deferred` |
+| A_ESCALATE 3-strike | ask_question 3-way | auto INCONCLUSIVE |
+| A_DISCOVER hit routing | ask_question | auto create issue |
+| A_DISCOVER ambiguous items | ask_question | all `deferred` |
+
+`deferred` items shown in completion summary; recoverable via `-c`.
+
+### Goal Prompt convergence rules
+
+```
+Exhaust iteration until all findings actioned (fix/issue/decision)
+and phase_goals_all_done=true.
+Fix by severity tiers, re-verify after each tier.
+Baseline captured before fix, compared after to confirm improvement.
+Pending decisions must ask_question — no silent resolve.
+```
+
+</appendix>
+
+</state_machine>
+
+<error_codes>
+| Code | Severity | Condition | Recovery |
+|------|----------|-----------|----------|
+| E001 | error | No target specified | Provide target or use -c |
+| E002 | error | Target path not found | Check path |
+| W001 | warning | No dependency manifest found | Proceed without dep audit |
+| W002 | warning | Some dimension agents failed | Partial audit coverage |
+| W003 | warning | Generalization 0 hits after full 3-layer scan | Advance to S_RECORD (requires all 3 layers attempted with evidence) |
+</error_codes>
+
+<success_criteria>
+- [ ] Target resolved, baseline metrics captured
+- [ ] Survey + 6-dimension audit with structured findings and severity matrix
+- [ ] Root causes diagnosed for critical/high findings
+- [ ] Improvements implemented and verified with before/after metrics (unless --skip-fix)
+- [ ] Multi-layer generalization + cross-phase loops (unless --skip-generalize)
+- [ ] Every unfixed finding has individual classification and reason
+- [ ] understanding.md §8 (metrics) and §9 (learnings) completed
+- [ ] phase_goals G1-G7 tracked and audited
+- [ ] Session resumable via -c
+- [ ] Completion summary
+</success_criteria>
+
+<next_step_routing>
+| Condition | Next |
+|-----------|------|
+| Security findings need deep investigation | `/odyssey-debug "<finding>"` |
+| UI-related findings | `/odyssey-ui "<component>"` |
+| Issues created from discoveries | `/manage-issue list --source improve-odyssey` |
+| Architecture pattern to document | `/spec-add arch "..."` |
+| Performance pattern to persist | `/spec-add coding "..."` |
+| Formal review of changes | `/odyssey-review-test-fix <changed-files>` |
+| Pending decisions | Filter evidence phase=decision status=pending |
+</next_step_routing>
+</output>