Token导航 LogoToken导航

Agent Skill · 研究检索

code-review(代码审查)

code-review 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。

code-review 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,权限为需要联网;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 53,999。

总安装 517周安装 22GitHub Stars 3下载 181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:code-review(代码审查)
来源仓库:https://github.com/odyssey4me/agent-skills
仓库路径:skills/code-review
安装命令:
npx skills add https://github.com/odyssey4me/agent-skills --skill code-review
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/odyssey4me/agent-skills --skill code-review

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。

code-review 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。适用于代码审查、质量检查和最佳实践查询等研究检索场景。

通过 npx skills add 命令从指定 GitHub 仓库安装,需结合原始 README 核验具体用法和功能边界。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操作。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/code-review

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/odyssey4me/agent-skills --skill code-review;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

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

code-review 怎么安装?

在终端执行 npx skills add https://github.com/odyssey4me/agent-skills --skill code-review。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

code-review 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。

code-review 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

code-review 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。

code-review 用的人多吗?

公开数据是累计安装 517 次,近一周安装 22 次,下载 181 次,GitHub 3 星。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。

code-review 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。

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。

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
code-review

代码审查 · 本页

5173需要联网未明确有提醒

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

Code Review

Orchestrates code review across GitHub PRs, GitLab MRs, and Gerrit changes. Auto-detects the platform from git remote configuration and provides focused review feedback on security, maintainability, and architectural fit.

This is a workflow skill -- it contains no scripts and instead guides the agent through a multi-step review process using the appropriate platform skill.

Authentication

This skill delegates authentication to the underlying platform skill:

  • GitHub: Requires gh auth login (see the github skill)
  • GitLab: Requires glab auth login (see the gitlab skill)
  • Gerrit: Requires git-review configuration (see the gerrit skill)

Ensure the relevant platform skill is authenticated before using code-review.

Commands

review

Review a change by number or URL.

Usage:

Review PR #123
Review this MR: https://gitlab.com/org/repo/-/merge_requests/42
Review Gerrit change 456789

The agent follows the Workflow steps: detects the platform from git remotes or the provided URL, fetches the change metadata, CI status, and diff, then provides structured review feedback. Optionally posts review comments.

remember

Save additional context for the current repository's reviews. This persists information that should be considered in future reviews of the same repo.

Usage:

Remember that this repo follows the Google Python Style Guide
Remember: authentication changes must be reviewed by the security team
Remember https://internal-docs.example.com/api-conventions as a reference for API design
Remember that the data layer uses the Repository pattern, not Active Record

Keyword: The word remember at the start of a message triggers saving. The context is stored in ~/.config/agent-skills/code-review.yaml under the current repository's remote URL.

What to save: Coding standards, architectural decisions, external documentation links, team conventions, review policies, or any context that should inform future reviews.

forget

Remove previously saved context for the current repository.

Usage:

Forget the note about the Google Python Style Guide
Forget all saved context for this repo

show context

Display all saved context for the current repository.

Usage:

Show review context for this repo

check

Verify that the required platform skill is available and authenticated.

# For GitHub repos
skills/github/scripts/github.py check

# For GitLab repos
skills/gitlab/scripts/gitlab.py check

# For Gerrit repos
skills/gerrit/scripts/gerrit.py check

Repository Context

Per-repository context is persisted in ~/.config/agent-skills/code-review.yaml, keyed by the remote fetch URL from git remote get-url origin. This context is loaded at the start of every review (see Step 0 in Workflow).

# ~/.config/agent-skills/code-review.yaml
repositories:
  "git@github.com:myorg/myrepo.git":
    references:
      - "https://internal-docs.example.com/api-conventions"
      - "https://google.github.io/styleguide/pyguide.html"
    standards:
      - "All API endpoints must validate input with Pydantic models"
      - "Authentication changes require security team review"
    notes:
      - "Data layer uses Repository pattern, not Active Record"
      - "Legacy modules in src/compat/ are exempt from new style rules"
  "https://gitlab.com/myorg/other-repo.git":
    references:
      - "https://docs.example.com/other-repo/architecture"
    standards: []
    notes:
      - "Migrating from REST to GraphQL -- new endpoints should use GraphQL"

When the user provides out-of-repo context during a review, suggest using the remember command to persist it.

Workflow

Step 0: Load Repository Context

Before starting the review, check for saved context:

git remote get-url origin

Read ~/.config/agent-skills/code-review.yaml and look up the remote URL. If context exists, load it and keep it in mind throughout the review:

  • references: Consult these when evaluating architectural decisions
  • standards: Actively check compliance with each standard
  • notes: Factor these into review feedback

If no context file exists or the repo has no entries, proceed without additional context.

Step 1: Detect Platform

Determine the code hosting platform from the repository context:

# Check git remotes
git remote -v
  • If remote contains github.com -> use the github skill
  • If remote contains gitlab -> use the gitlab skill
  • If .gitreview file exists -> use the gerrit skill
  • If a URL is provided, detect from the URL hostname

Step 2: Fetch Change Metadata and CI Status

GitHub:

skills/github/scripts/github.py prs view <number> --repo OWNER/REPO
skills/github/scripts/github.py prs checks <number> --repo OWNER/REPO

GitLab:

skills/gitlab/scripts/gitlab.py mrs view <number> --repo GROUP/REPO
skills/gitlab/scripts/gitlab.py pipelines list --repo GROUP/REPO

Gerrit:

skills/gerrit/scripts/gerrit.py changes view <change-number>

Step 3: Assess CI/Test Status

Before reviewing, check whether CI/tests have passed:

  • If CI is passing: proceed with full review
  • If CI is failing: note the failures, skip reviewing concerns that would be caught by tests, and focus on issues tests cannot catch (security, architecture, design)
  • If CI is pending: note it and proceed with review

Step 4: Fetch the Diff

GitHub:

gh pr diff <number>

GitLab:

glab mr diff <number>

Gerrit:

git diff HEAD~1

Step 5: Review the Changes

Focus review feedback on these areas, in priority order. See references/review-checklist.md for the full checklist.

  1. Security concerns: injection vulnerabilities, authentication/authorization gaps, data exposure, unsafe deserialization, hardcoded secrets
  2. Maintainability: excessive complexity, poor naming, missing separation of concerns, code duplication that harms readability
  3. Good coding practices: error handling gaps, resource leaks, race conditions, missing input validation at system boundaries
  4. Architectural fit: consistency with existing codebase patterns, appropriate abstraction level, dependency direction

Do not flag:

  • Style/formatting issues (leave to linters)
  • Minor naming preferences without clear readability impact
  • Test coverage gaps (leave to CI coverage tools)
  • Issues already caught by failing CI

Step 6: Present Findings

Format findings as a structured review:

## Code Review: PR #<number> - <title>

### Summary
<1-2 sentence summary of the change and overall assessment>

### CI Status
<passing/failing/pending -- note any failures>

### Findings

#### Security
- [ ] <finding with file:line reference>

#### Maintainability
- [ ] <finding with file:line reference>

#### Coding Practices
- [ ] <finding with file:line reference>

#### Architecture
- [ ] <finding with file:line reference>

### Verdict
<APPROVE / REQUEST_CHANGES / COMMENT -- with brief rationale>

If the user requests it, post the review as comments on the PR/MR using the platform skill:

GitHub:

gh pr review <number> --comment --body "<review>"
# Or approve/request changes:
gh pr review <number> --approve --body "<review>"
gh pr review <number> --request-changes --body "<review>"

GitLab:

glab mr note <number> --message "<review>"
# Or approve:
glab mr approve <number>

Examples

Review a GitHub PR

Review PR #42

The agent will run git remote -v, detect GitHub, fetch the PR with skills/github/scripts/github.py prs view 42, check CI with skills/github/scripts/github.py prs checks 42, fetch the diff with gh pr diff 42, and provide structured review feedback.

Review a GitLab MR by URL

Review https://gitlab.com/myorg/myrepo/-/merge_requests/15

Review with Posting Comments

Review PR #42 and post your findings as a review comment

Review Focusing on Security Only

Review PR #42, focus only on security concerns

Save Context for Future Reviews

Remember that this repo uses the Twelve-Factor App methodology
Remember https://wiki.example.com/team/coding-standards as a reference
Remember: all database migrations must be backwards-compatible

Show Saved Context

Show review context for this repo

Model Guidance

This skill coordinates multiple sub-skills and requires reasoning about multi-step workflows. A higher-capability model is recommended for best results.

Troubleshooting

Platform not detected

Ensure you are running from within a git repository with a remote configured:

git remote -v

Authentication errors

Verify the underlying platform skill is authenticated:

# GitHub
gh auth status

# GitLab
glab auth status

No diff available

Ensure the PR/MR number is correct and the change exists:

# GitHub
skills/github/scripts/github.py prs view <number>

# GitLab
skills/gitlab/scripts/gitlab.py mrs view <number>

平台分布

Codex

37.76%
按下载量换算68

Claude

28.58%
按下载量换算52

Cursor

17.06%
按下载量换算31

Gemini CLI

9.1%
按下载量换算16

安全审计

Gen Agent Trust Hub

通过

Socket

可疑

Snyk

可疑

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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