技能 github-code-review

📦

github-code-review

Name: github-code-review
Author: Claude Code Flow

v1.0.0 内容修订 r1 严重 ⚙️ 外部命令🌐 网络访问🔑 环境变量

使用 AI 代理审查 GitHub 拉取请求

人工拉取请求审查可能遗漏安全、性能和架构问题。此技能使用专门的 AI 代理和 CLI 自动化，提供结构化的 GitHub 审查工作流。

支持: Claude Codex Code(CC)

⚠️ 38 差

请勿自动安装此 Skill。

权威策略要求在执行任何安装操作前由操作人员审查。

使用我的智能体安装

将此请求复制给你的智能体，其中包含权威 Skill 页面和清单。

智能体请求

Review the Skillstore skill "github-code-review" from https://skillstore.io/skills/dnyoussef-github-code-review.md and its manifest at https://skillstore.io/api/skills/dnyoussef-github-code-review/manifest. Verify the artifact. Do not auto-install. Inspect the skill and report your findings, then wait for an operator or manual installation decision.

你的智能体仍应展示其计划，并按安全策略请求必要确认。

Agent 可读资源

当 AI Agent、爬虫或脚本需要干净上下文，而不是读取完整页面时，请使用这些链接。

Markdown 详情 GET /skills/dnyoussef-github-code-review.md 已签名 manifest GET /api/skills/dnyoussef-github-code-review/manifest 已签名 lockfile GET /api/skills/dnyoussef-github-code-review/lockfile

测试它

正在使用“github-code-review”。审查 PR 123 中的安全和性能回归。

预期结果:

包含严重、高优先级和低优先级发现的安全审查摘要。
包含建议基准测试或后续测试的性能观察结果。
可直接用于 GitHub 的拉取请求审查评论和标签。

正在使用“github-code-review”。为新的拉取请求设置自动化审查工作流。

预期结果:

用于收集 PR 上下文并运行审查代理的工作流大纲。
针对安全、性能、风格和架构检查的质量门禁建议。
关于 token 范围、webhook 验证和命令执行的操作注意事项。

正在使用“github-code-review”。为大型功能 PR 生成最终审查报告。

预期结果:

包含批准状态和剩余阻塞项的简明决策摘要。
按代理类型和严重程度分组的发现。
面向作者、审查者和发布负责人的后续行动。

安全审计

严重

v10 • 2026/7/9 打开版本化报告

The skill is a GitHub pull request review workflow that intentionally uses gh, npx, shell commands, and GitHub tokens. Most Markdown backtick findings are false positives, but webhook-driven execSync, authenticated automation, external command execution, and automated approval or merge examples are real risks that require controls before publication.

已扫描文件

1,141

分析行数

复核项

已忽略误报

已确认安全问题 (2)

严重

SKILL.md:376-407

Untrusted PR Comment Command Execution

The skill presents PR comments such as /swarm commands as direct automation triggers and shows a webhook handler passing the comment body into a shell command. This can let untrusted repository users influence privileged review automation.

The command path is explicit: PR comments are accepted when they start with /swarm, then the full comment body is passed to execSync. Later checklist guidance does not remove the unsafe example.

高

SKILL.md:528-535 SKILL.md:996-1008 SKILL.md:1035-1038

Automated Approval and Merge Workflow Risk

The skill includes examples that approve pull requests based on generated review output and enable auto-merge after review counts. That can bypass human judgment if adopted without strict policy controls.

The cited examples issue gh pr review --approve and gh pr merge commands from automation flows. They are documented examples, but the business logic risk is clear if used directly.

能力复核项 (26)

这些是真实的本地能力，对此技能可能是预期行为，因此需要复核，但不会按已确认恶意行为计分。

严重

SKILL.md:407

Synchronous process execution

execSync(`npx ruv-swarm github handle-comment --pr ${event.issue.number} --command "${command}"`);

The command embeds a PR comment body into execSync after only checking that it starts with /swarm. Untrusted comment text can influence shell execution.

严重

SKILL.md:407-414

Ruby/shell backtick execution

execSync(`npx ruv-swarm github handle-comment --pr ${event.issue.number} --command "${command}"`);

The backtick appears inside an execSync command template in the webhook example. The pattern is executable JavaScript command construction, not just documentation formatting.

高

SKILL.md:395

Node.js child_process module

const { execSync } = require('child_process');

The webhook example imports child_process to run local commands from GitHub webhook events. That enables privileged command execution if copied into production without strict validation.

高

SKILL.md:402

Synchronous process execution

execSync(`npx ruv-swarm github pr-init ${event.pull_request.number}`);

The webhook example synchronously runs an npx ruv-swarm command from event data. This is real command execution in response to webhook input and needs strong trust controls.

高

SKILL.md:402-407

Ruby/shell backtick execution

execSync(`npx ruv-swarm github pr-init ${event.pull_request.number}`);

The backtick appears inside an execSync command template in the webhook example. The pattern is executable JavaScript command construction, not just documentation formatting.

高

SKILL.md:512

Git platform tokens

run: echo "${{ secrets.GITHUB_TOKEN }}" | gh auth login --with-token

The workflow authenticates gh with secrets.GITHUB_TOKEN before running review automation. This is legitimate but exposes repository privileges to the surrounding automation path.

中

SKILL.md:39

Shell command substitution

PR_DATA=$(gh pr view 123 --json files,additions,deletions,title,body)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:40

Shell command substitution

PR_DIFF=$(gh pr diff 123)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:107

Shell command substitution

PR_DATA=$(gh pr view 123 --json files,additions,deletions,title,body)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:108

Shell command substitution

PR_DIFF=$(gh pr diff 123)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:138

Shell command substitution

CHANGED_FILES=$(gh pr view 123 --json files --jq '.files[].path')

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:141-145

Shell command substitution

SECURITY_RESULTS=$(npx ruv-swarm github review-security \

The shell substitution runs npx ruv-swarm and captures its output for later GitHub actions. This executes external tooling with PR context and should require explicit trust.

中

SKILL.md:518

Shell command substitution

PR_DATA=$(gh pr view $PR_NUM --json files,title,body,labels)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:519

Shell command substitution

PR_DIFF=$(gh pr diff $PR_NUM)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:522-526

Shell command substitution

REVIEW_OUTPUT=$(npx ruv-swarm github review-all \

The shell substitution runs npx ruv-swarm and captures its output for later GitHub actions. This executes external tooling with PR context and should require explicit trust.

中

SKILL.md:557

Shell command substitution

PR_DIFF=$(gh pr diff 123 --color never)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:558

Shell command substitution

PR_FILES=$(gh pr view 123 --json files)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:561-567

Shell command substitution

COMMENTS=$(npx ruv-swarm github review-comment \

The shell substitution runs npx ruv-swarm and captures its output for later GitHub actions. This executes external tooling with PR context and should require explicit trust.

中

SKILL.md:571

Shell command substitution

FILE=$(echo "$comment" | jq -r '.path')

The shell substitution parses generated review output for later API calls. The pattern is operationally legitimate but depends on untrusted or tool-generated data.

中

SKILL.md:572

Shell command substitution

LINE=$(echo "$comment" | jq -r '.line')

The shell substitution parses generated review output for later API calls. The pattern is operationally legitimate but depends on untrusted or tool-generated data.

中

SKILL.md:573

Shell command substitution

BODY=$(echo "$comment" | jq -r '.body')

The shell substitution parses generated review output for later API calls. The pattern is operationally legitimate but depends on untrusted or tool-generated data.

中

SKILL.md:574

Shell command substitution

COMMIT_ID=$(gh pr view 123 --json headRefOid -q .headRefOid)

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

中

SKILL.md:814

Shell command substitution

PROGRESS=$(npx ruv-swarm github pr-progress 123 --format markdown)

The shell substitution runs npx ruv-swarm and captures its output for later GitHub actions. This executes external tooling with PR context and should require explicit trust.

中

SKILL.md:819

Shell command substitution

if [[ $(echo "$PROGRESS" | grep -o '[0-9]\+%' | sed 's/%//') -gt 90 ]]; then

The shell substitution parses generated review output for later API calls. The pattern is operationally legitimate but depends on untrusted or tool-generated data.

中

SKILL.md:1000

Shell command substitution

SWARM_STATUS=$(npx ruv-swarm github pr-status 123)

The shell substitution runs npx ruv-swarm and captures its output for later GitHub actions. This executes external tooling with PR context and should require explicit trust.

中

SKILL.md:1004

Shell command substitution

REVIEWS=$(gh pr view 123 --json reviews --jq '.reviews | length')

The shell substitution invokes gh to fetch pull request data from GitHub. This is expected for the skill but still performs authenticated external access.

风险因素

⚙️ 外部命令 (134)

SKILL.md:395 SKILL.md:402 SKILL.md:407 SKILL.md:28-34 SKILL.md:34-37 SKILL.md:37-49 SKILL.md:49-105 SKILL.md:105-120 SKILL.md:120-136 SKILL.md:136-156 SKILL.md:156-161 SKILL.md:161-180 SKILL.md:180-187 SKILL.md:187-199 SKILL.md:199-201 SKILL.md:201-206 SKILL.md:206-216 SKILL.md:216-223 SKILL.md:223-228 SKILL.md:228-246 SKILL.md:246-256 SKILL.md:256-263 SKILL.md:263-268 SKILL.md:268-286 SKILL.md:286-296 SKILL.md:296-302 SKILL.md:302-307 SKILL.md:307-325 SKILL.md:325-335 SKILL.md:335-345 SKILL.md:345-351 SKILL.md:351-362 SKILL.md:362-366 SKILL.md:366-372 SKILL.md:372-380 SKILL.md:380-387 SKILL.md:387-392 SKILL.md:392-402 SKILL.md:402-407 SKILL.md:407-414 SKILL.md:414-424 SKILL.md:424-459 SKILL.md:459-463 SKILL.md:463-486 SKILL.md:486-494 SKILL.md:494-547 SKILL.md:547-555 SKILL.md:555-585 SKILL.md:585-589 SKILL.md:589-596 SKILL.md:596-604 SKILL.md:604-614 SKILL.md:614-618 SKILL.md:618-628 SKILL.md:628-632 SKILL.md:632-639 SKILL.md:639-649 SKILL.md:649-657 SKILL.md:657-663 SKILL.md:663-676 SKILL.md:676-682 SKILL.md:682-689 SKILL.md:689-693 SKILL.md:693-699 SKILL.md:699-707 SKILL.md:707-757 SKILL.md:757-761 SKILL.md:761-767 SKILL.md:767-775 SKILL.md:775-798 SKILL.md:798-802 SKILL.md:802-808 SKILL.md:808-812 SKILL.md:812-822 SKILL.md:822-830 SKILL.md:830-838 SKILL.md:838-842 SKILL.md:842-850 SKILL.md:850-854 SKILL.md:854-862 SKILL.md:862-866 SKILL.md:866-873 SKILL.md:873-877 SKILL.md:877-884 SKILL.md:884-892 SKILL.md:892-897 SKILL.md:897-901 SKILL.md:901-908 SKILL.md:908-912 SKILL.md:912-918 SKILL.md:918-922 SKILL.md:922-928 SKILL.md:928-976 SKILL.md:976-994 SKILL.md:994-998 SKILL.md:998-1011 SKILL.md:1011-1027 SKILL.md:1027-1052 SKILL.md:1052-1064 SKILL.md:1064-1073 SKILL.md:1073-1081 SKILL.md:1081-1090 SKILL.md:1090-1098 SKILL.md:1098-1107 SKILL.md:1107-1116 SKILL.md:1116-1117 SKILL.md:1117-1118 SKILL.md:39 SKILL.md:40 SKILL.md:107 SKILL.md:108 SKILL.md:138 SKILL.md:141-145 SKILL.md:518 SKILL.md:519 SKILL.md:522-526 SKILL.md:557 SKILL.md:558 SKILL.md:561-567 SKILL.md:571 SKILL.md:572 SKILL.md:573 SKILL.md:574 SKILL.md:814 SKILL.md:819 SKILL.md:1000 SKILL.md:1004 SKILL.md:37-49 SKILL.md:105-120 SKILL.md:136-156 SKILL.md:494-547 SKILL.md:555-585 SKILL.md:812-822 SKILL.md:998-1011

🌐 网络访问 (3)

SKILL.md:1121 SKILL.md:1122 SKILL.md:1123

🔑 环境变量 (1)

SKILL.md:512

审计者: codex 查看审计历史 →

分享与引用此报告

分享版本化评估报告、中性徽章、嵌入卡片和引用。Skillstore 只报告证据，不判断该 Skill 是否安全。

打开版本化报告

复制报告链接

https://skillstore.io/skills/dnyoussef-github-code-review/audits/10?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown 徽章

[![Skillstore security assessment](https://skillstore.io/badges/skills/dnyoussef-github-code-review/security.svg)](https://skillstore.io/skills/dnyoussef-github-code-review?utm_source=security_passport_badge)

HTML 徽章

<a href="https://skillstore.io/skills/dnyoussef-github-code-review?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/dnyoussef-github-code-review/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

嵌入卡片

<iframe src="https://skillstore.io/embed/skills/dnyoussef-github-code-review.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

学术引用 (APA · BibTeX · CFF)

APA 引用

Claude Code Flow. (2026). github-code-review security audit report (audit version 10) [Author version 1.0.0]. Skillstore. https://skillstore.io/skills/dnyoussef-github-code-review/audits/10

BibTeX 引用

@techreport{claude-code-flow-dnyoussef-github-code-review-2026,
  author = {Claude Code Flow},
  title = {github-code-review security audit report (audit version 10)},
  institution = {Skillstore},
  year = {2026},
  number = {10},
  url = {https://skillstore.io/skills/dnyoussef-github-code-review/audits/10},
  note = {Author version 1.0.0}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "github-code-review security audit report (audit version 10)"
version: "1.0.0"
type: report
authors:
  - name: "Claude Code Flow"
date-released: "2026-07-09"
url: "https://skillstore.io/skills/dnyoussef-github-code-review/audits/10"
identifiers:
  - type: other
    value: "skillstore:dnyoussef-github-code-review:audit:10"
    description: "Skillstore immutable audit report identifier"

比较变体

2 个可安装变体

每位作者的技能都保持为独立的可安装项。推荐变体按 Skillstore 证据排序。

为什么此变体排名第一

Skillstore 使用量更高

作者 / 来源 Skillstore 评分证据置信度 Skillstore 使用量已更新

ruvnet 推荐

ruvnet-github-code-review

Skillstore 评分 38

证据置信度中

Skillstore 使用量 13

已更新

2026-09-09

Claude Code Flow 当前

dnyoussef-github-code-review

Skillstore 评分 38

证据置信度中

Skillstore 使用量 9

已更新

2026-09-09

Skillstore 评分

评分依据证据置信度: 中

架构

100

可维护性

内容

社区

规范符合性

你能构建什么

审查安全关键型拉取请求

在批准前，针对身份验证、授权、密钥和依赖项更改运行聚焦的审查工作流。

标准化团队审查工作流

使用可重复的审查代理和质量门禁，提高各个拉取请求之间的一致性。

自动化 PR 反馈收集

将 GitHub Actions、审查评论、标签和指标连接成一个统一的审查流程。

试试这些提示

运行基础 PR 审查

使用 GitHub 代码审查技能审查拉取请求 123。重点关注正确性、测试和清晰反馈。

聚焦安全发现

审查拉取请求 123 中的安全问题。优先关注身份验证、授权、密钥、注入风险和依赖项更改。

创建多代理审查计划

为拉取请求 123 创建一个审查计划，使用安全、性能、架构、风格和可访问性代理。包括预期输出和质量门禁。

审计审查自动化工作流

评估拉取请求 123 的自动化审查工作流。检查 webhook 安全性、token 范围、命令验证、批准规则和合并控制。

最佳实践

仅在可信仓库中，并在获得明确用户批准且 GitHub token 范围受限的情况下运行命令。
在任何自动化执行前，验证 PR 评论、webhook 签名和命令参数。
将生成的审查用作决策支持，并对高风险变更保留人工批准。

避免

不要允许任意 PR 评论触发 shell 命令或特权自动化。
不要将未经审查的模型输出直接传递给 GitHub 批准、合并或打标签命令。
除非 package 和版本可信，否则不要在包含仓库 secrets 的情况下运行外部 npm packages。

常见问题

此技能是否需要 GitHub 访问权限？

是。大多数工作流需要 gh CLI 身份验证以及用于拉取请求数据和评论的仓库权限。

它可以自动批准或合并拉取请求吗？

该技能包含批准和合并自动化示例，但团队应要求这些操作经过人工审查。

它涵盖哪些审查领域？

它涵盖安全、性能、架构、风格、可访问性、测试、质量门禁和审查指标。

将其用于公开拉取请求是否安全？

请谨慎使用。公开拉取请求可能包含不可信输入，因此 tokens、webhooks 和命令需要严格控制。

它会安装或运行外部工具吗？

示例使用 gh、npx、ruv-swarm、claude-flow、jq、npm 和 shell 命令。使用前请验证每个工具。

它可以替代人工审查者吗？

不能。它有助于组织审查工作并生成反馈，但最终决策应由人工审查者做出。

开发者详情

作者

Claude Code Flow

许可证

MIT

作者版本

v1.0.0

Skillstore 修订

仓库

https://github.com/DNYoussef/ai-chrome-extension/tree/main/.claude/skills/github-code-review

引用

d29bab7c799689e19147d92070e306a76db0a8a9

维护新鲜度

2026/7/21

使用量

5 次下载 · 343 次浏览

文件结构

📄 SKILL.md

Claude Code Flow 的更多技能

📦

slash-command-encoder

为 AI 工作流创建斜杠命令

当用户依赖冗长的自然语言提示时，重复性的 AI 工作流会很慢。此技能可帮助定义简洁的斜杠命令，并支持路由、验证、链式调用和帮助文本。

效率查看

📦

micro-skill-creator

创建聚焦的 AI 微技能

可复用的 AI 工作流常常会变得过于宽泛且难以测试。此技能可帮助定义聚焦的技能，并配备专家提示词、契约和验证计划。

效率查看

📦

cascade-orchestrator

编排多模型技能级联

复杂的 AI 工作通常需要多种技能、模型和质量检查。此技能可帮助设计包含路由、sandbox 测试、swarm 协调和记忆的级联。

编码与开发查看

📦

web-cli-teleport

选择 Claude Code Web 或 CLI

当团队为任务选择了错误的 Claude Code 界面时，会浪费时间。此技能会推荐 Web、CLI 或混合交接工作流。

效率查看

📦

agentdb-semantic-vector-search

构建 AgentDB 语义搜索

团队需要为 RAG 和知识库提供相关的文档检索。此技能指导 AgentDB 设置、嵌入存储、索引、API 搜索和相关性调优。

数据与分析查看

📦

when-creating-presentations-use-pptx-generation

创建无障碍 PowerPoint 演示文稿

手动制作演示文稿可能缓慢且不一致。此技能会引导 Claude、Codex 或 Claude Code 完成结构化 PPTX 生成和验证。

办公与报告查看

查看全部

📦

code-reviewer

使用结构化质量检查审查代码

由 alirezarezvani

代码审查可能会漏掉质量、安全性和可维护性问题。此技能通过脚本、检查清单和报告生成来指导结构化审查。

编码与开发查看

📦

gauntlet

Review Defects and Verify Root-Cause Repairs

由 aakashh242

Reviews can miss failure conditions or accept fixes without reliable evidence. Gauntlet guides scoped adversarial reviews, authorized repairs, and revision-bound verification within a finite budget.

编码与开发查看

📦

creating-issues

创建一致的 GitHub Issue

由 C0ntr0lledCha0s

当不同仓库中的 issue 标题、标签和项目元数据不一致时，团队会浪费时间。此 skill 可指导 Claude、Codex 和 Claude Code 创建一致的 GitHub issue。

编码与开发查看

📦

nestjs-code-audit

Audit NestJS Code with Evidence-Backed Findings

由 amirtaherkhani

NestJS repositories can hide boundary flaws and runtime risks that lint results alone cannot reveal. This skill combines local evidence collection and semantic review into one prioritized report.

编码与开发查看

📦

building-agents

构建 Claude Code 智能体

由 C0ntr0lledCha0s

创建可靠的 AI 智能体需要清晰的 schema、安全的工具和可重复的审查步骤。此技能为 Claude Code 工作流中的智能体设计、验证、迁移和维护提供指导。

编码与开发查看

📦

building-commands

构建 Claude Code 斜杠命令

由 C0ntr0lledCha0s

团队需要易于调用和审查的可重复 AI 工作流。此技能可帮助创建、更新、验证和记录 Claude Code 斜杠命令。

编码与开发查看

github-code-review

使用我的智能体安装

Agent 可读资源

测试它

安全审计

已确认安全问题 (2)

风险因素

复制报告链接

Markdown 徽章

HTML 徽章

嵌入卡片

APA 引用

BibTeX 引用

CITATION.cff

比较变体

Skillstore 评分

你能构建什么

审查安全关键型拉取请求

标准化团队审查工作流

自动化 PR 反馈收集

试试这些提示

最佳实践

避免

常见问题

开发者详情

Claude Code Flow 的更多技能

slash-command-encoder

micro-skill-creator

cascade-orchestrator

web-cli-teleport

agentdb-semantic-vector-search

when-creating-presentations-use-pptx-generation

相关技能

code-reviewer

gauntlet

creating-issues

nestjs-code-audit

building-agents

building-commands