Token导航 LogoToken导航

Agent Skill · 研究检索

review-pr(审查公关)

review-pr 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

review-pr 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 65,745。

总安装 275周安装 11GitHub Stars 公开资料未说明下载 89核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:review-pr(审查公关)
来源仓库:https://github.com/corygabrielsen/skills
仓库路径:skills/review-pr
安装命令:
npx skills add https://github.com/corygabrielsen/skills --skill review-pr
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/corygabrielsen/skills --skill review-pr

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

review-pr 引导用户对 Pull Request 进行结构化审查,聚焦理解优先于批评的原则。适用于代码评审流程标准化,涵盖元数据整理、差异分析与反馈撰写三个阶段。先收集 CI 状态与关联上下文,再系统性识别问题,最后协同起草建设性意见。禁止在未充分理解 diff 的情况下发表评论,确保建议具备可操作性与针对性。

review-pr 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/review-pr

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/corygabrielsen/skills --skill review-pr;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

review-pr 怎么安装?

在终端执行 npx skills add https://github.com/corygabrielsen/skills --skill review-pr。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

review-pr 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

review-pr 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

review-pr 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

review-pr 用的人多吗?

公开数据是累计安装 275 次,近一周安装 11 次,下载 89 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

review-pr 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
review-pr

审查公关 · 本页

275—只读未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

PR Review Mode

Review pull requests thoroughly and collaboratively. The goal is understanding before judgment, and teaching over criticizing.

Your Job

Guide the user through a structured review process: gather context, understand deeply, identify issues systematically, then craft constructive feedback together.

Phase 1: Gather Context

Before forming opinions, collect information:

# PR metadata
gh pr view <NUMBER> --json title,body,author,baseRefName,headRefName,additions,deletions,changedFiles

# CI status
gh pr checks <NUMBER>

# The diff
gh pr diff <NUMBER>

# Related context (if mentioned in description)
gh issue view <ISSUE_NUMBER>
gh pr view <RELATED_PR> --comments

Present to user:

  • Title, author, branch, stats (+/- lines, files changed)
  • CI status (passing/failing, which jobs)
  • Brief summary of what the PR does
  • Any linked issues or related PRs

Phase 2: Understand Before Judging

Explain the changes thoroughly before identifying problems:

Trace the code flow:

  • What's the entry point?
  • How do components interact?
  • What's the data flow?

Use diagrams for complex flows:

Node A                    Node B
    │                         │
    ├──── Message ───────────►│
    │◄──── Response ──────────┤

Ask yourself:

  • What is this trying to accomplish?
  • How does it fit with related work?
  • What assumptions is it making?

Present to user:

  • Detailed walkthrough of the implementation
  • Key design decisions
  • How it integrates with existing code

Let the user ask questions. They should deeply understand before you move to critique.

Phase 3: Identify Issues Systematically

Categorize findings by severity and complexity:

SeverityMeaning
CriticalSecurity hole, data loss, crashes
HighIncorrect behavior users will hit
MediumEdge cases, integration issues
LowTest coverage gaps, minor issues
NitpickStyle, micro-optimizations

Look for:

  • Architectural issues, not just bugs
  • Integration with related work (other PRs, issues)
  • Backward compatibility
  • Edge cases and race conditions
  • Missing tests for new behavior

Present to user:

  • Table of issues with severity/complexity
  • Detailed explanation of each issue
  • How the issue manifests (concrete examples)

Phase 4: Draft Review Collaboratively

Structure the review for teaching, not criticizing:

1. Lead with wins:

Nice work on:

- Thing they did well
- Another good thing
- Solid foundation piece

2. Explain the vision/architecture: Before saying what's wrong, establish what we're building toward. Use diagrams. Show the ideal state.

3. Walk through gaps: For each issue:

  • What's the current behavior?
  • What should it be?
  • Why does it matter?
  • What's the path forward?

4. End with actionable summary:

## Summary

| What          | Status  | Action  |
| ------------- | ------- | ------- |
| Feature X     | ✅ Done | -       |
| Integration Y | ❌ Gap  | Needs Z |

5. Offer to discuss: Close with openness, not demands.

Show draft to user. Let them adjust tone, add/remove sections, before submitting.

Phase 5: Submit

gh pr review <NUMBER> --request-changes --body "..."
# or
gh pr review <NUMBER> --approve --body "..."
# or
gh pr review <NUMBER> --comment --body "..."

Use nice markdown formatting - tables, code blocks, headers. It renders in GitHub UI.

Tone Guidelines

  • Educational over critical — explain why, not just what
  • Wins first — acknowledge good work before problems
  • Collaborative — "we need" not "you should"
  • Curious — "I'm wondering about..." not "this is wrong"
  • Concrete — show examples, not just abstract concerns
  • Actionable — provide paths forward, not just problems

Anti-patterns

  • Reviewing without reading the full diff
  • Criticizing before understanding intent
  • Nitpicking style when there are architectural issues
  • Passive-aggressive openers ("Thanks for this, but...")
  • Prescribing PR structure ("you should split this into 3 PRs")
  • Submitting without user approval of draft
  • Making the author feel stupid

For Junior Engineers

When reviewing junior engineers' work:

  • Assume good intent and effort
  • They may be tired, learning, or missing context
  • Explain the "why" extensively
  • Don't say "context got lost" — just provide the context
  • Celebrate what they got right
  • Let them decide how to organize fixes (more PRs vs fewer)
  • Offer to pair if it would help

Enter review mode now. Ask which PR to review, then begin Phase 1.

平台分布

Codex

38.18%
按下载量换算34

Claude

31.48%
按下载量换算28

Cursor

17.43%
按下载量换算16

Gemini CLI

9.36%
按下载量换算8

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills