requesting-code-review skill
✓ VerifiedDispatches a specialized sub-agent to perform structured code reviews by providing specific git SHAs and task descriptions instead of full session history.
- GitHub stars
- 8,529
- Files
- 2
- Last commit
- Nov 11, 2026
- License
- MIT
- Allowed tools
- —
What does the requesting-code-review skill do?
requesting-code-review dispatches a specialized sub-agent to perform structured code reviews by providing specific git SHAs and task descriptions instead of full session history.
Summary generated from the skill's SKILL.md.
About the requesting-code-review skill
Developers who want to validate their work before merging code can use this skill to dispatch a specialized sub-agent for structured reviews. Instead of overwhelming a reviewer with entire session histories, the tool provides a focused context using specific git SHAs and task descriptions. This approach keeps the reviewer focused on the actual code changes rather than the developer's thought process, preserving the primary agent's context window.
An agent should load this skill when a task is finished, a major feature is implemented, or right before merging into the main branch. The process requires the user to provide a base SHA and a head SHA along with a description of the work and the original requirements. The skill utilizes a general-purpose task tool and follows a template found in code-reviewer.md to produce organized feedback categorized by severity levels like critical, important, and minor.
What an agent can do with it
- 1Verify completed tasksRequest a code review immediately after finishing a specific task in a sub-agent driven development workflow.
- 2Prepare for merging codePerform a structured review of changes before merging them into the main branch.
- 3Establish refactoring baselinesUse a code review to create a baseline before starting a refactoring process.
- 4Resolve development blocksDispatch a reviewer to get a fresh perspective when you are stuck on a problem.
Written from the skill's SKILL.md. Every use case is checked against a quote from it.
How to install the requesting-code-review skill
Each command copies the skill folder into the client's skills directory. Restart the client or start a new session to load it.Install the requesting-code-review skill in Plugin
/plugin marketplace add jnMetaCode/superpowers-zh /plugin install superpowers-zh@superpowers-zh
Install the requesting-code-review skill in Claude Code
~/.claude/skills/requesting-code-review/SKILL.md
$ npx degit jnMetaCode/superpowers-zh/skills/requesting-code-review ~/.claude/skills/requesting-code-reviewInstall the requesting-code-review skill in Claude Code (project)
.claude/skills/requesting-code-review/SKILL.md
$ npx degit jnMetaCode/superpowers-zh/skills/requesting-code-review .claude/skills/requesting-code-reviewInstall the requesting-code-review skill in Codex
~/.agents/skills/requesting-code-review/SKILL.md
$ npx degit jnMetaCode/superpowers-zh/skills/requesting-code-review ~/.agents/skills/requesting-code-reviewInstall the requesting-code-review skill in Gemini CLI
~/.gemini/skills/requesting-code-review/SKILL.md
$ npx degit jnMetaCode/superpowers-zh/skills/requesting-code-review ~/.gemini/skills/requesting-code-reviewInstall the requesting-code-review skill in Claude.ai
npx degit jnMetaCode/superpowers-zh/skills/requesting-code-review requesting-code-review && zip -r requesting-code-review.zip requesting-code-review # claude.ai → Customize → Skills → + → Create skill → Upload a skill → requesting-code-review.zip
Automated checks
Ran Nov 15, 2026Full report →Passed our automated checks. That is not a security audit: read the scripts before you let an agent run them. How checks work · What to fix for more points
- Source repository linked and reachablejnmetacode/superpowers-zh/skills/requesting-code-review is public and reachablegithub.com/jnmetacode/superpowers-zh/tree/HEAD/skills/reques ↗Pass
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit 2026-11-11github.com/jnmetacode/superpowers-zh/commits ↗Pass
- No download-and-run or credential theft in scriptsNo scripts in the skill folderPass
- No hidden instructions in SKILL.mdNo override phrases, safeguard switches, credential requests or hidden charactersgithub.com/jnmetacode/superpowers-zh/blob/HEAD/skills/reques ↗Pass
- Pre-approved tools are scopedNo pre-approved tools; the agent asks before using onegithub.com/jnmetacode/superpowers-zh/blob/HEAD/skills/reques ↗Pass
- SKILL.md follows the Agent Skills formatthe description does not say when to use the skill; the description is only 32 charactersgithub.com/jnmetacode/superpowers-zh/blob/HEAD/skills/reques ↗Finding
SKILL.md
View on GitHub请求代码审查
派遣代码审查子代理,在问题扩散之前发现它们。审查者获得的是精心组织的评估上下文——绝不是你的会话历史。
核心原则: 早审查,勤审查。
何时请求审查
必须审查:
- 子代理驱动开发中每个任务完成后
- 完成重要功能后
- 合并到 main 之前
可选但有价值:
- 卡住时(换个视角)
- 重构之前(建立基线)
- 修复复杂 bug 之后
如何请求
1. 获取 git SHA:
BASE_SHA=$(git rev-parse HEAD~1) # 或:git merge-base origin/main HEAD
HEAD_SHA=$(git rev-parse HEAD)
2. 派遣代码审查子代理:
使用 Task 工具,指定 general-purpose 类型,填写 code-reviewer.md 中的模板
占位符说明:
{DESCRIPTION}- 你刚完成的内容简要说明{PLAN_OR_REQUIREMENTS}- 预期功能{BASE_SHA}- 起始提交{HEAD_SHA}- 结束提交
3. 处理反馈:
- Critical 问题立即修复
- Important 问题在继续之前修复
- Minor 问题记录下来稍后处理
- 如果审查者有误,用技术理由反驳
示例
[刚完成任务 2:添加验证功能]
你:让我在继续之前请求代码审查。
BASE_SHA=$(git log --oneline | grep "Task 1" | head -1 | cut -d' ' -f1)
HEAD_SHA=$(git rev-parse HEAD)
[派遣代码审查子代理]
DESCRIPTION: 添加了 verifyIndex() 和 repairIndex(),支持 4 种问题类型
PLAN_OR_REQUIREMENTS: docs/superpowers/plans/deployment-plan.md 中的任务 2
BASE_SHA: a7981ec
HEAD_SHA: 3df7661
[子代理返回]:
优点:架构清晰,测试真实
问题:
Important:缺少进度指示器
Minor:报告间隔使用了魔法数字 (100)
评估:可以继续
你:[修复进度指示器]
[继续任务 3]
常见的合理化借口
| 借口 | 现实 |
|---|---|
| "我自己看一下 diff 就行了,不用专门派审查者" | 你是协调者——在自己的会话里读 diff 会烧掉你继续推进工作所需的上下文窗口。派一个审查子智能体:diff 和评估过程都待在它的上下文里,只有结论回到你这里。 |
| "审查者需要我的全部会话历史才能理解这次改动" | 给它精心组织的上下文,绝不给会话历史。这样审查者才会盯着工作成果,而不是你的思考过程。 |
红线
绝不要:
- 因为"很简单"就跳过审查
- 忽略 Critical 问题
- 带着未修复的 Important 问题继续推进
- 对合理的技术反馈进行争辩
如果审查者有误:
- 用技术理由反驳
- 展示证明其可行的代码/测试
- 要求澄清
参见模板:requesting-code-review/code-reviewer.md
Files in this skill
2requesting-code-review skill questions
How do I install the requesting-code-review skill in Claude Code?
Run npx degit jnMetaCode/superpowers-zh/skills/requesting-code-review ~/.claude/skills/requesting-code-review. Claude Code loads skills from ~/.claude/skills (all projects) or .claude/skills (one project) and uses a skill when its description matches the task.
Which clients can use the requesting-code-review skill?
Any client that reads the Agent Skills format: Claude Code, Codex, Gemini CLI and claude.ai (upload as a ZIP). Scripts in the skill run only where the client allows code execution.
When should I request a code review?
You must request a review after completing each task in sub-agent driven development, after finishing important features, or before merging to main.
Should I provide my entire session history to the reviewer?
No, you should provide organized assessment context such as descriptions and git SHAs, but never the full session history to keep the reviewer focused on the work output.
How should I handle different levels of feedback from the reviewer?
Fix Critical issues immediately, fix Important issues before continuing, and record Minor issues to be addressed later.