skill-review-response
Use when a reviewer, CI bot, or another AI leaves feedback to address
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- skill-review-response
- description
- Use when a reviewer, CI bot, or another AI leaves feedback to address
- disable-model-invocation
- true
02 资源文件 1
- 参考文档agents/openai.yaml176 B
03 SKILL.md 正文
Host: Codex CLI — This skill was designed for Claude Code and adapted for Codex. Cross-reference commands use installed skill names in Codex rather than
/octo:*slash commands. Use the active Codex shell and subagent tools. Do not claim a provider, model, or host subagent is available until the current session exposes it. For host tool equivalents, seeskills/blocks/codex-host-adapter.md.
Receiving Code Review
Core Principle
Code review requires technical evaluation, not performative agreement.
Never blindly implement review feedback. Verify it's correct for THIS codebase before changing anything.
The Response Pattern
WHEN receiving code review feedback:
1. READ — Complete feedback without reacting
2. RESTATE — Summarize the requirement in your own words
3. VERIFY — Check against actual codebase state
4. EVALUATE — Is this technically sound for THIS context?
5. RESPOND — Technical acknowledgment OR reasoned pushback
6. IMPLEMENT — One item at a time, verify each change
Forbidden Responses
NEVER say:
- "You're absolutely right!" (without verification)
- "Great catch!" (before confirming it IS a catch)
- "I'll fix that right away!" (before evaluating whether it needs fixing)
- "Done!" (without running verification — see skill-verification-gate)
These are social performance, not technical evaluation. They lead to:
- Implementing wrong suggestions
- Introducing bugs to "fix" non-issues
- Wasting time on style preferences disguised as bugs
Evaluation Checklist
For each piece of feedback:
| Question | If YES | If NO |
|---|---|---|
| Is the issue real? (verify in code) | Continue evaluation | Push back with evidence |
| Does the suggested fix work here? | Continue evaluation | Propose alternative |
| Does fixing this break something else? | Fix both or push back | Implement the fix |
| Is this a style preference or a real problem? | Acknowledge, deprioritize | Fix it |
| Was this already considered and rejected? | Explain the trade-off | Implement |
How to Push Back
When feedback is wrong or doesn't apply:
> Reviewer: "This function should handle null input"
>
> Response: "Checked — this function is only called from `processUser()`
> (line 47) which validates non-null before dispatch. Adding null handling
> here would be dead code. The caller contract guarantees non-null."
Provide:
- What you checked
- Why the suggestion doesn't apply
- Evidence (line numbers, call sites, tests)
Multi-Provider Review Context
In Claude Octopus workflows, review feedback comes from multiple sources:
- Codex review — tends toward enterprise patterns, may over-engineer
- Antigravity review — tends toward ecosystem conformity, may suggest unnecessary dependencies
- Claude review — tends toward elegance, may under-engineer error handling
- Sonnet review — tends toward thoroughness, may flag low-priority issues
When providers disagree:
- Check which provider's suggestion matches the ACTUAL codebase conventions
- The codebase's existing patterns win over any provider's preferences
- If two providers flag the same issue, it's probably real
Handling Feedback Loops
When a reviewer flags an issue and you fix it:
- Make the fix
- Run verification (skill-verification-gate) — prove the fix works
- Re-read the original feedback — did you address the root cause or just the symptom?
- If the reviewer re-reviews and finds new issues, that's normal — don't get frustrated
- Each round should have FEWER issues, not different ones
If the same issue keeps coming back:
- You're fixing symptoms, not the root cause
- Stop and re-read the feedback from scratch
- Ask the reviewer to clarify if the issue is ambiguous
When Review Feedback Conflicts with Requirements
If a reviewer suggests something that contradicts the spec/requirements:
- Note the conflict explicitly
- Check if the spec is wrong (it might be)
- If spec is correct: implement the spec, note the reviewer's concern for future consideration
- If spec is wrong: flag to the user before changing anything
Requirements trump review suggestions. User intent trumps both.
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.