| name | production-quality-gatekeeper |
| description | Tự động thiết lập và thực thi vòng lặp tự phản biện và hoàn thiện (self-refining loop) cho AI Agent đạt chuẩn Production-grade. |
| disable-model-invocation | true |
| user-invocable | true |
=== BOOT CONFIGURATION (L0 — Anchor Rules) ===
must:
- run the core_case.py status check at startup before taking any actions
- load the appropriate domain knowledge (creative, dev, or llm) based on the target task
- run scripts/loop_refiner.py on the draft after every turn to programmatically verify quality
- read the generated .skill-context/production-quality-gatekeeper/feedback.yaml upon loop failure (exit 1)
- perform highly targeted incremental edits focused exclusively on the failed criteria
- iterate up to 10 times until loop_refiner.py returns exit 0 (100% score)
- generate the final 📊 Production Quality Evaluation Report using the template on success
must_not:
- skip the validation loop or assume success without running loop_refiner.py
- compromise on minor or major failed criteria
- rewrite passed sections during refinement, to avoid context bloat and code regression
### Boot Sequence
1. Read `SKILL.md` (this file) — done
2. Read `../_shared/knowledge/framework.md` — Stage boundaries and conventions
3. Read `../_shared/knowledge/case-system.md` — CASE System specifications
4. Run `python3 ../_shared/validators/check_status.py .skill-context/{skill-name}/design.md` to verify current status.
- If checkpoint stale (> 7 days), warn user.
5. Identify the target domain (creative, dev, or llm) and load the corresponding knowledge document.
6. Proceed to Phase 1: Quality Criteria Synthesis
Progressive Disclosure Plan
