Skip to main content

code-review-assistant

Analyzes PRs to assign expert reviewers, detect code issues, and generate architectural diagrams. Use for PR reviews, code analysis, pull requests, git diffs, or mentions of reviewers, code quality, or standards.

来源信息

仓库
pbakaus/code-review-assistant
最近来源活动
2025年11月24日 02:13
检测到的 SKILL.md 语言
英语
星标
10
Forks
4

安装方式

默认使用会先检查来源的 Prompt;你也可以切换为直接命令,或下载本地副本。

检查来源文件

决定是否安装前,请先阅读 SKILL.md,以及 SkillsMP 当前展示的配套文件。

文件资源管理器
11 个文件

正在显示 SKILL.md

SKILL.md
来源说明 · 只读预览
name
code-review-assistant
description
Analyzes PRs to assign expert reviewers, detect code issues, and generate architectural diagrams. Use for PR reviews, code analysis, pull requests, git diffs, or mentions of reviewers, code quality, or standards.
allowed-tools
["Bash","Read","Write","Grep","Glob","WebFetch"]
# Code Review Assistant Analyzes a pull request to assign reviewers, detect code issues, and visually explain complex changes. Can automatically assign reviewers via GitHub CLI and post review comments. **Setup**: Run `bash scripts/setup.sh` to check environment and install dependencies. ## Get PR Data **If you have a diff and PR link provided:** Analyze that directly. **Otherwise:** Use GitHub CLI to fetch from current repo: ```bash gh pr view --json number,title,body,author,files gh pr diff ``` If `gh` not installed: https://cli.github.com/ ## Analysis 1. **Assign Reviewers**: Read `reference/team-expertise.md`, match expertise areas to changes in the diff 2. **Detect Issues**: - Read `reference/code-standards-map.md` for quick issue list - If any patterns match the diff, grep the full details from `code-standards.md` using the XML tag - Flag by severity (🔴🟡🔵) 3. **Visual Explanations**: Unless none or only minor code changes, run: - `node scripts/generate-diagram.js "<description>"` - `node scripts/upload-image.js <file>` → Cloudinary URL - Include as `![](url)` in review ### Output Format ```markdown ## 🤖 Claude PR Assessment ## 🎨 Visual Explanation ![](url) - Description (don't include if minimal code changes) ## 👥 Recommended Reviewers - **@username** (Name) - Expertise areas that match this PR ## ⚠️ Areas of Concern - 🔴/🟡/🔵 **Matched code standard pattern title** - Files it occurs in, optional 3-5 line snippets and aggregated explanation - Concise recommended solution ## ✅ Overall Assessment Strengths, changes needed, recommendation ``` ## Post-Review Actions After presenting the review, offer to help with: 1. **Auto-assign reviewers**: ```bash gh pr edit <number> --add-reviewer username1,username2,username3 ``` Use the GitHub usernames (@username) from the recommended reviewers list. 2. **Post review as PR comment**: ```bash gh pr comment <number> --body-file <review-file.md> ``` Save the full review markdown to a temporary file first, then post it. Ask the user if they'd like you to perform either of these actions. ## Reference Files Progressive disclosure - only load what you need: - `reference/team-expertise.md` - Team member expertise areas with GitHub usernames - `reference/code-standards-map.md` - Quick list of detectable issues (check this first) - `reference/code-standards.md` - Full issue details with XML tags (grep specific sections) ## Scripts - `scripts/generate-diagram.js` - Gemini API integration for visual explanations - `scripts/upload-image.js` - Cloudinary upload for public image URLs Run with Node.js, see script headers for usage details.
在 GitHub 查看