| name | co-review |
| description | Layer 3 promote-candidate review for claude-codex-orchestration. Every 5 sessions (or when .eval-scores.jsonl has 5+ new entries), scan the weak-point + root-cause distribution in recent logs, identify recurring patterns (≥ 2× occurrences), score each candidate on Durability + Impact + Scope (0–3 each), flag those ≥ 6 for /co-promote. Delegates to claude-codex-orchestration references/self-correction.md Layer 3. |
/co-review — Promote-Candidate Review (Layer 3)
Full protocol: ~/.claude/skills/claude-codex-orchestration/references/self-correction.md §Layer 3.
Quick run
- Read
.eval-scores.jsonl + .error-log.jsonl in the skill directory
- Group by
weak_point and root_cause
- For each recurring pattern (appears ≥ 2 times):
- Check if SKILL.md / references/ already addresses it
- Score: Durability + Impact + Scope (each 0–3)
- ≥ 6 → promote candidate (list for
/co-promote)
- 4–5 → watch
- ≤ 3 → ignore
- Output ordered list of promote candidates with reasoning
- Suggest user run
/co-promote <candidate> for the top one
Also check .issue-log.jsonl — if an external issue (Layer 2.5) can be closed by promoting this candidate, note that in output.