Requesting Code Review
SkillDocs & knowledgeUse this skill at Step 8 of the v2 SOP after Verify passes, to request and receive a code review from a Reviewer-Agent (or human reviewer). Packages the diff with context (task documents, specs, COMPLIANCE_TAG, design rationale), invokes the review, and iterates with systematic-debugging until all feedback is closed. Chinese trigger examples: "请求 review", "Step 8", "代码审查", "请人 review", "review 这个 change". Do NOT use as substitute for Step 10 compliance review, do NOT skip when reviewer feedback contains High items. Success = all reviewer feedback closed, review thread archived.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Requesting Code Review skill
What this skill tells your AI
The instructions your AI receives, as published by charliehzm/medharness in .claude/skills/requesting-code-review/SKILL.md and read by ahel’s review.
The "second pair of eyes" gate. Functional, not compliance — that's Step 10.
Review packet contents
When asking for a review, always include:
- The diff (limited to the task's ≤ 2 files)
- The task document
- The relevant spec id(s)
- Reference to design.md and (if applicable) ARCH_INPUT_INDEX
- The PHI handling note (if L3/L4)
What the reviewer checks
| Layer | Examples |
|---|---|
| Correctness | matches spec; covers acceptance criteria |
| Code health | readable, lean (no dead code, no extra comments) |
| Security | input validation, error handling at boundaries, injection surface |
| Maintainability | tests align, naming sensible |
| Reuse | did the author miss an existing utility (reference/legacy-utilities-index.md)? |
Workflow
- Compose review packet.
- Dispatch to Reviewer-Agent (with heterogeneous model — DON'T use coder's model) or to a human.
- Capture feedback in
REVIEW_THREAD.md. - For each feedback item: classify Critical / Major / Minor.
- Address Critical + Major via
systematic-debugging. - Minor: address or accept-and-record.
- Close when all Critical and Major are closed.
Common failure modes
- Self-review on same model — coder and reviewer share model = correlated blind spots. Mitigation: enforce heterogeneity.
- Drive-by approvals — reviewer skims and approves. Mitigation: require explicit checklist signoff.
- Long-running review threads — drags. Mitigation: 24h SLA; auto-escalate.
Signals
- GitHub stars
- 86
- Forks
- 8
- Last commit
- Jun 2026
Advanced
- Catalog kind
- skill
- Gateway key
requesting-code-review-charliehzm- Source
- github.com/charliehzm/medharness