菜单
Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Use when the user says "ask grok", "get grok's take", "grok review", "have grok look at this", "delegate to grok", or any variation naming Grok/xAI as the perspective they want.
Use when starting any creative work - creating a feature, building a component, adding functionality, or changing designed behavior - before writing code or invoking any implementation skill. Small defect repairs and tweaks triage through razorback:fixing-small-issues first.
Use when a design question resists discussion — a state model with more edge cases than fit in your head, a UI the user keeps flip-flopping on because nobody can picture it, or behavior only knowable by running it — before writing a design doc or implementation plan for that question.
Use when the user says "ask claude", "fresh claude review", "second opinion from another claude", "have another claude look at this", "delegate to a fresh claude", or any variation naming Claude as the second perspective they want.
Use when the user says "ask codex", "get codex's take", "codex review", "have codex look at this", "delegate to codex", or any variation naming Codex/OpenAI as the perspective they want. Also use for a generic "second opinion from a different model" when no other model is named.
Use when planning or reviewing non-trivial code changes, refactoring architecture, evaluating module boundaries, repeated findings reveal coupling, tests are hard to write because interfaces are unclear, or the user asks for codebase design improvements, complexity reduction, deduplication, or cleanup of existing code.
基于 SOC 职业分类
| name | receiving-code-review |
| description | Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable |
Code review requires technical evaluation, not emotional performance.
Core principle: Verify before implementing. Ask before assuming. Technical correctness over social comfort.
WHEN receiving code review feedback:
1. READ: Complete feedback without reacting
2. UNDERSTAND: Restate requirement in own words (or ask)
3. VERIFY: Check against codebase reality
4. EVALUATE: Technically sound for THIS codebase?
5. RESPOND: Technical acknowledgment or reasoned pushback
6. IMPLEMENT: One item at a time, test each
NEVER:
INSTEAD:
Unclear external-review items during an approved autonomous run (the common case): verify the clear items, fix independently verifiable real issues, flag the unclear item for review and continue — noting it in the report — unless it matches the blocker taxonomy. Do not let one vague external suggestion block unrelated safe fixes, and do not stop the run to ask about it.
Unclear human direction blocks implementation (interactive review from the user, outside an approved run): ask first, because partial understanding can send the whole fix in the wrong direction.
IF any item from the user is unclear (interactive):
STOP - do not implement anything yet
ASK for clarification on unclear items
WHY: Items may be related. Partial understanding = wrong implementation.
Example (interactive):
User: "Fix 1-6"
You understand 1,2,3,6. Unclear on 4,5.
❌ WRONG: Implement 1,2,3,6 now, ask about 4,5 later
✅ RIGHT: "I understand items 1,2,3,6. Need clarification on 4 and 5 before proceeding."
BEFORE implementing:
1. Check: Technically correct for THIS codebase?
2. Check: Breaks existing functionality?
3. Check: Reason for current implementation?
4. Check: Works on all platforms/versions?
5. Check: Does reviewer understand full context?
6. Check: Inspect the symbol being discussed with Miller `inspect(target, depth=overview)` — escalate to `depth=full` for the symbol the feedback centers on
7. Check: Find references to verify suggested changes won't break callers with Miller `trace`
IF suggestion seems wrong:
Push back with technical reasoning
IF can't easily verify:
Investigate with Miller and the smallest relevant verification command.
If still not verifiable during an approved autonomous run, classify it as flagged-for-review in the report and continue unless it matches the blocker taxonomy.
If this is an interactive review outside an approved run, ask one specific clarifying question.
IF conflicts with the user's prior decisions:
In an approved autonomous run, keep the prior decision, flag the conflict in the report, and continue unless it matches the blocker taxonomy.
Outside an approved run, ask one specific question before overriding the prior decision.
External architecture feedback is evaluated through razorback:architecture-quality before implementation.
Rule: external feedback - be skeptical, but check carefully.
IF reviewer suggests "implementing properly":
Find references with Miller `trace` to check actual usage across codebase
IF unused: "This endpoint isn't called. Remove it (YAGNI)?"
IF used: Then implement properly
Rule: the reviewer doesn't set scope, the user does. If the feature isn't needed, don't add it.
FOR multi-item feedback:
1. Clarify anything unclear FIRST
2. Then implement in this order:
- Blocking issues (breaks, security)
- Simple fixes (typos, imports)
- Complex fixes (refactoring, logic)
3. Test each fix individually
4. Verify no regressions
Push back when:
How to push back:
When feedback IS correct:
✅ "Fixed. [Brief description of what changed]"
✅ "Good catch - [specific issue]. Fixed in [location]."
✅ [Just fix it and show in the code]
❌ "You're absolutely right!"
❌ "Great point!"
❌ "Thanks for catching that!"
❌ "Thanks for [anything]"
❌ ANY gratitude expression
Why no thanks: Actions speak. Just fix it. The code itself shows you heard the feedback.
If you catch yourself about to write "Thanks": DELETE IT. State the fix instead.
If you pushed back and were wrong:
✅ "You were right - I checked [X] and it does [Y]. Implementing now."
✅ "Verified this and you're correct. My initial understanding was wrong because [reason]. Fixing."
❌ Long apology
❌ Defending why you pushed back
❌ Over-explaining
State the correction factually and move on.
| Mistake | Fix |
|---|---|
| Performative agreement | State requirement or just act |
| Blind implementation | Verify against codebase first |
| Batch without testing | One at a time, test each |
| Assuming reviewer is right | Check if breaks things |
| Avoiding pushback | Technical correctness > comfort |
| Partial implementation | Clarify all items first |
| Can't verify, proceed anyway | Investigate, then classify under blocker taxonomy or flag for review |
Performative Agreement (Bad):
Reviewer: "Remove legacy code"
❌ "You're absolutely right! Let me remove that..."
Technical Verification (Good):
Reviewer: "Remove legacy code"
✅ "Checking... build target is 10.15+, this API needs 13+. Need legacy for backward compat. Current impl has wrong bundle ID - fix it or drop pre-13 support?"
YAGNI (Good):
Reviewer: "Implement proper metrics tracking with database, date filters, CSV export"
✅ "Grepped codebase - nothing calls this endpoint. Remove it (YAGNI)? Or is there usage I'm missing?"
Unclear Item (Good, interactive):
User: "Fix items 1-6"
You understand 1,2,3,6. Unclear on 4,5.
✅ "Understand 1,2,3,6. Need clarification on 4 and 5 before implementing."
When replying to inline review comments on GitHub, reply in the comment thread (gh api repos/{owner}/{repo}/pulls/{pr}/comments/{id}/replies), not as a top-level PR comment.
External feedback = suggestions to evaluate, not orders to follow.
Verify. Question. Then implement.
No performative agreement. Technical rigor always.