receiving-code-review
Teaches how to receive review feedback properly — verify-before-implement on every comment, different skepticism levels by source, evidence-based pushback, no performative agreement. Prevents blindly
SKILL.mdskills/receiving-code-review/SKILL.md
GitHubReceiving Code Review
Problem it solves
Agents (and humans) often handle review feedback poorly:
- Blindly agreeing with every comment without understanding it
- Implementing suggestions without verifying they're correct
- Performative agreement — saying "good point, fixed" when the fix doesn't address the issue
- Getting defensive or dismissing valid feedback
- Wasting effort on changes the reviewer didn't actually request
- Circular review loops from misunderstanding feedback
Protocol
1. Understand before acting
For each review comment:
- Read it completely — don't skim
- Paraphrase it back to yourself: "They want me to change X because Y"
- Ask if unclear — if the comment is ambiguous, ask: "Just to confirm, you want X changed to Y, correct?"
Do NOT start implementing until you understand what's being asked and why.
2. Verify-before-implement
Before making any change requested in review:
- Is the suggestion correct? Will it actually solve the stated problem?
- Does it follow the project's patterns? (see follow-existing-patterns)
- Does it maintain the spec? (see spec-driven-development)
- Are there side effects? What else might break?
If the answer to any of these is "no" or "I'm not sure", do not implement — investigate first or push back.
3. Skepticism levels by source
Not all review feedback is equal. Apply skepticism based on source:
| Source | Skepticism | Approach |
|---|---|---|
| Author (you) | High | Self-review before submitting. You have the most context but also the most blind spots. |
| Domain expert | Low | Strong signal. Understand the rationale carefully. |
| Peer | Medium | Good signal. Verify against spec and patterns. |
| Non-domain expert | Medium-High | Check if comment is about correctness vs preference. |
| Automated tool | Low for rules, High for suggestions | Linters/rules are ground truth. Tool suggestions need human judgment. |
4. Evidence-based pushback
If you disagree with a review comment:
- Don't say "I disagree" — say "I think X is better because Y"
- Provide evidence: "The spec says Z", "This follows the existing pattern in file.rs:42"
- Offer alternatives: "Would approach A work instead of suggested B?"
- Accept if you're wrong: "You're right, I missed that edge case. Fixed."
5. No performative agreement
Never do these:
- Saying "fixed" when you haven't actually made the change
- Saying "good point" when you don't actually agree
- Making the change without understanding why
- Marking a comment resolved without verifying the fix works
Instead:
- "I see your point about X. I've updated the code to Y."
- "Good catch on the edge case. Added handling for it."
- "I understand the concern, but I think the current approach is correct because..."
6. Keep a change log
When making review-driven changes:
- Track each requested change and its resolution status
- Verify each change passes tests
- Re-read the diff before submitting the next round
Detection triggers
Activate when:
- Review feedback is received (from any source: human, AI, tool)
- Implementing changes based on someone else's suggestion
- About to respond to a review comment
- User says "fix this" and you need to evaluate if it's correct
When NOT to use
- You are the reviewer (use code-review instead)
- Clear bug fixes with unambiguous root cause (use debugging-and-error-recovery)
- User explicitly says "just do it, trust me"
Cross-references
code-review — Use together. code-review gives reviews; receiving-code-review teaches how to respond to them.
code-collaborate-qa — For conversational back-and-forth about code. receiving-code-review is for structured response to formal review.
anti-sycophancy — Performative agreement is a form of sycophancy. Verify before agreeing.
follow-existing-patterns — When implementing review feedback, check that the change follows existing patterns.
spec-driven-development — Review suggestions must stay within spec scope. If a suggestion expands scope, push back or re-spec.
verify-and-cite — Verify review claims before implementing. Don't assume the reviewer is right.
anti-premature-termination — Do not declare "all review comments addressed" until each is verified and changes are tested.