review-feedback-promoter
Convert repeated review findings into durable harness checks. Use when code review catches recurring agent mistakes that should become docs, tests, lint rules, boundary checks, validation scripts, or
SKILL.mdskills/review-feedback-promoter/SKILL.md
GitHubReview Feedback Promoter
Purpose
Turn recurring human or agent review feedback into reusable harness defenses instead of repeating the same comment.
Inspect First
- recent review findings,
agent-revieweroutput, quality docs, lint/boundary/test scripts, architecture docs, dependency/security policies, and known repeated defects
Procedure
Classify each finding.
- Documentation rule for knowledge gaps.
- Test for behavior gaps.
- Boundary or lint check for mechanical architecture gaps.
- Security/dependency/migration check for governance gaps.
- Entropy/golden-pattern update for consistency gaps.
Record promotions.
- Create or update
docs/quality/review-feedback-promotions.md. - Include source finding, promoted defense, validation command, owner, and status.
- Create or update
Add the lightest durable defense.
- Prefer existing project-native checks.
- Add scripts only when deterministic and repeatable.
- Avoid broad rules that create false positives.
Route specialized fixes.
- Use
architecture-boundary-enforcerfor import/layer issues. - Use
testing-strategy-builderortdd-playwright-workflowfor behavior gaps. - Use
documentation-sync-enforcer,security-baseline, ordependency-governancefor specialized policy gaps.
- Use
Validation
- Confirm each promotion maps to a real repeated finding.
- Confirm each new rule has a command, doc location, or test path.
- Confirm the defense is specific enough to avoid noisy failures.
Completion Criteria
- Repeated review feedback has a durable follow-up path.
- The harness grows stronger without bloating root instructions or adding unnecessary tools.