AI-Skills

English | 中文

A comprehensive library of modular skills for Claude Code. Each skill provides specialized AI assistance for common development tasks.

Features

AI-Skills includes 9 specialized skills covering the entire development lifecycle:

Skill	Description	Trigger With
commit-helper	Generate Conventional Commits messages	"create commit", "write commit message"
debug-helper	Systematic debugging framework	"debug this", "help troubleshoot"
code-explainer	Explain code with visual aids	"explain this code", "how does this work"
test-generator	Generate unit tests	"write tests", "add test coverage"
api-design-helper	Design RESTful APIs	"design an API", "create API endpoints"
documentation-helper	Write technical documentation	"help document this", "write docs"
logging-helper	Add structured logging	"add logging", "implement logging"
performance-optimizer	Optimize code performance	"optimize this", "improve performance"
pr-review-helper	Conduct PR code reviews	"review this PR", "code review"

Quick Start

Installation

Clone this repository:

git clone https://github.com/Youpen-y/AI-Skills.git
cd AI-Skills

Copy skills to your project:

# Copy all skills to your project
cp -r skills/* /path/to/your-project/.claude/skills/

# Or copy specific skill
cp -r skills/commit-helper /path/to/your-project/.claude/skills/

Copy skills globally (for all projects):

# Linux/macOS
mkdir -p ~/.claude/skills
cp -r skills/* ~/.claude/skills/

# Windows
mkdir %USERPROFILE%\.claude\skills
xcopy /E /I skills\* %USERPROFILE%\.claude\skills\

Using Skills

Skills are automatically discovered by Claude Code through the SKILL.md frontmatter. Simply use natural language or slash commands:

# Example: Create a commit
claude "commit these changes"

# Example: Debug an issue
claude "debug this error"

# Example: Write tests
claude "write tests for auth.js"

Project Structure

AI-Skills/
├── skills/                         # Skill definitions
│   ├── commit-helper/             # Conventional Commits
│   ├── debug-helper/              # Debugging framework
│   ├── code-explainer/            # Code explanation
│   ├── test-generator/            # Test generation
│   ├── api-design-helper/         # API design
│   ├── documentation-helper/      # Documentation
│   ├── logging-helper/            # Logging
│   ├── performance-optimizer/     # Performance
│   └── pr-review-helper/          # PR reviews
└── .github/workflows/
    ├── ai-pr-review.yml           # AI PR Review
    └── ai-pr-chat.yml             # AI PR Chat Bot

GitHub Actions Integration

The project includes two AI-powered workflows using Zhipu's GLM-4.7 model:

AI PR Review

Automated code reviews that trigger on PR creation/update:

• Fetches PR diff via GitHub API
• Sends to GLM-4.7 for analysis
• Posts review with structured feedback
• Updates existing bot comments to avoid duplicates

Setup: Add ZHIPU_API_KEY to your repository secrets.

AI PR Chat Bot

Interactive AI assistant for PR comments:

• Triggered by /ai or @ai mentions
• Detects "review mode" vs "question mode"
• Provides concise, focused responses
• Supports follow-up questions

Skill Architecture

Each skill follows a standardized structure:

skill-name/
├── SKILL.md           # Core skill definition (required)
├── README.md          # User documentation (required)
├── examples.md        # Real-world examples (optional)
├── reference.md       # Detailed specifications (optional)
└── scripts/           # Executable implementations (optional)

SKILL.md Format

---
name: skill-name
description: When to use this skill. Includes trigger patterns.
---

# Skill Name

Usage instructions and guidelines...

Contributing

Contributions are welcome! To add a new skill:

1. Create a directory under /skills/<skill-name>/
2. Add SKILL.md with YAML frontmatter
3. Create a comprehensive README.md
4. Optionally add examples.md and reference.md
5. Follow the existing skill structure

Commit Standards

This project follows Conventional Commits 1.0.0:

<type>[optional scope]: <description>

[optional body]

[optional footer(s)]

Types: feat, fix, docs, style, refactor, perf, test, build, ci, chore, revert

License

MIT

Acknowledgments

Built for Claude Code by Anthropic.

AI workflows powered by Zhipu AI GLM-4.7 model.