odyssey-review-test-fix

A_INTAKE

Parse target + flags → file list. Create SESSION_DIR, derive phase_goals[]. Search prior knowledge. Write session.json + section 1. Call create_goal with phase_goals as success_criteria.

A_ARCHAEOLOGY

spawn_agents_on_csv (Wave 1):

"arch-timeline","Git Timeline","git log --oneline -20 -- {target_files}","archaeology","","","1","pending","","",""
"arch-blame","Git Blame","git blame on key regions of target files","archaeology","","","1","pending","","",""

spawn_agents_on_csv({ csv_path: "tasks.csv", id_column: "id",
  instruction: ARCHAEOLOGY_INSTRUCTION + TERMINATION_CONTRACT,
  max_concurrency: 2, max_runtime_seconds: 300,
  output_csv_path: "wave-1-results.csv", output_schema: SHARED_OUTPUT_SCHEMA })

Merge → evidence (phase: archaeology). CLI delegate --to claude --mode analysis. Update section 2.

A_EXPLORE

CLI delegate --to claude --mode analysis — call chains, error gaps, similar patterns. Write explore.json. Update section 3. Mark G2.

A_REVIEW

spawn_agents_on_csv (Wave 2):

"rev-correct","Correctness","Logic errors, boundary conditions, null/undefined, race conditions","review","correctness","","2","pending","","",""
"rev-security","Security","Injection, XSS, CSRF, data exposure, auth bypass","review","security","","2","pending","","",""
"rev-perf","Performance","Hot paths, N+1, memory leaks, unnecessary recomputation","review","performance","","2","pending","","",""
"rev-arch","Architecture","Layer violations, circular deps, interface contracts, SoC","review","architecture","","2","pending","","",""

spawn_agents_on_csv({ csv_path: "tasks.csv", id_column: "id",
  instruction: REVIEW_INSTRUCTION + TERMINATION_CONTRACT,
  max_concurrency: 4, max_runtime_seconds: 600,
  output_csv_path: "wave-2-results.csv", output_schema: SHARED_OUTPUT_SCHEMA })

Each returns [{title, severity, file, line, description, suggestion, cwe}]. Merge → evidence (review). Write review_result + section 4 (severity matrix). Mark G1.

A_FIX

Exhaustive tier loop — descend severity until remaining_actionable == 0:

for tier in [critical, high, medium, low].filter(>= threshold):
  for each unfixed candidate: read +/-20 → fix → evidence (fix)
  re-review modified area: new findings → append, continue (max 2/tier)
  tier done → auto-commit

Normal: request_user_input per tier. -y: auto-fix all. Remaining > 0 → retry (no max_loops limit). Unchanged 2 rounds → classify each individually. Blanket "pre-existing" forbidden.

A_CONFIRM

Run tests + CLI delegate zero-residual review (--to claude --mode analysis).

• remaining == 0 AND new == 0 → confirmed, mark G3
• Otherwise → needs_rework → S_FIX

Update confirmation + remaining_actionable + section 5.

A_GENERALIZE

Base shared_actions. Pattern source: findings (severity >= medium).

spawn_agents_on_csv (Wave 3):

"gen-syntax","Syntax Grep","Grep syntax-layer patterns across project","generalization","syntax","","3","pending","","",""
"gen-semantic","Semantic Scan","Check related modules for same anti-patterns","generalization","semantic","","3","pending","","",""
"gen-structural","Structural Match","Find structurally similar files, check for same issues","generalization","structural","","3","pending","","",""
"gen-historical","Historical Grep","git log -S pattern for introduction/fix history","generalization","historical","","3","pending","","",""

spawn_agents_on_csv({ csv_path: "tasks.csv", id_column: "id",
  instruction: GENERALIZATION_INSTRUCTION + TERMINATION_CONTRACT,
  max_concurrency: 4, max_runtime_seconds: 600,
  output_csv_path: "wave-3-results.csv", output_schema: SHARED_OUTPUT_SCHEMA })

Cross-layer dedup: multi-layer hits → boost confidence | single-layer → needs_review | historically fixed → regression_risk. Iterative deepening: module with >= 3 hits → targeted deep scan (max 1 round). Mark G4.

A_DISCOVER

Base shared_actions. Review overrides: cross-phase loop tracking per base.

A_RECORD

Base shared_actions. Learnings per Knowledge Persistence table. Confirmation gate: Before writing spec entries, present proposed entries to user via request_user_input for confirmation. Skip confirmation only if -y flag is set.

Completion summary:

--- REVIEW-TEST-FIX ODYSSEY COMPLETE ---
Target:     {target}          Dimensions: {dims}
Findings:   {C}C {H}H {M}M {L}L    Fix: {fixed}, confirmed={yes|skip}
Patterns:   {N} ({by_layer})        Scan hits: {total} ({cross} cross-layer)
Issues: {N}  Decisions: {N}r/{M}p/{K}d  Learnings: {N}  Self-iter: {N}x{M}
Goals:  {done}/{total} ({skipped} skipped)
---