Content
🚀 GLM MCP Skills
<div align="center">
**Transforming Zhipu MCP capabilities into lightweight Agent Skills**
[](LICENSE)
[](https://www.python.org/)
English
</div>
---
## English
### Motivation
Zhipu's GLM Coding Plan offers powerful MCP (Model Context Protocol) support, but native MCP has drawbacks in real-world usage (Claude Code, Gemini CLI, custom Agents):
- **Verbose Context**: Complete tool definitions required for each call
- **High Token Cost**: Protocol overhead significantly increases expenses
- **Complex Integration**: Requires handling MCP handshakes and sessions
This project transforms MCP capabilities into **lightweight, intuitive Skills**:
| Feature | Description |
|---------|-------------|
| **🎯 Context Efficiency** | Optimized prompts, reduced overhead, saved tokens |
| **🧩 Modular Skills** | Atomic capabilities inspired by Claude Code |
| **🔌 Easy Integration** | Quick integration with mainstream AI coding tools |
### Available Skills
| Skill | Description |
|-------|-------------|
| [**web-reader**](skills/web-reader/SKILL.md) | Extract main content from any web page |
| [**web-search**](skills/web-search/SKILL.md) | Real-time web search for latest information |
| [**zread**](skills/zread/SKILL.md) | Browse GitHub repos - structure, files, issues |
| [**zai-vision**](skills/zai-vision/SKILL.md) | AI image/video analysis - OCR, UI-to-code, diagnostics |
### Quick Start
#### 1. Clone
```bash
git clone https://github.com/eject-dream/glm-mcp-skills.git
```
#### 2. Copy Skills to Your Project
```bash
# Copy all skills
cp -r glm-mcp-skills/skills /path/to/your/project/
# Or copy a specific skill
cp -r glm-mcp-skills/skills/web-search /path/to/your/project/skills/
```
#### 3. Install Dependencies
```bash
cd glm-mcp-skills
uv sync # recommended
# or
pip install -r requirements.txt
```
#### 4. Configure Environment Variables
```bash
# Required: set API Key
export ZAI_API_KEY=your_api_key
# Optional: choose API provider (ZAI or ZHIPU), default ZHIPU
export Z_AI_MODE=ZHIPU
```
### Environment Variables
| Variable | Required | Default | Description |
|----------|----------|---------|-------------|
| `ZAI_API_KEY` | ✅ | - | Zhipu AI / Z.AI API Key |
| `Z_AI_MODE` | ❌ | `ZHIPU` | Provider: `ZAI` or `ZHIPU` |
| `ZAI_BASE_URL` | ❌ | - | Custom API URL (for local models) |
| `ZAI_VISION_MODEL` | ❌ | `glm-4.6v` | Vision model name |
| `ZAI_TIMEOUT` | ❌ | `300` | Request timeout (seconds) |
### Detailed Usage of Skills
#### 🔍 web-search - Web Search
```bash
# Search for latest news
python skills/web-search/scripts/web_search.py "latest tech news"
# Search for specific topics
python skills/web-search/scripts/web_search.py "Python 3.13 new features"
```
#### 📖 web-reader - Web Reader
```bash
# Read article content
python skills/web-reader/scripts/web_reader.py "https://example.com/article"
# Read technical documentation
python skills/web-reader/scripts/web_reader.py "https://docs.python.org/3/"
```
#### 📂 zread - GitHub Repository Browser
```bash
# Search repository documentation
python skills/zread/scripts/zread.py search vitejs/vite "configuration guide" zh
# View repository structure
python skills/zread/scripts/zread.py tree anthropics/anthropic-cookbook
# Read specific file
python skills/zread/scripts/zread.py read vercel/next.js package.json
```
#### 👁️ zai-vision - AI Vision Analysis
```bash
# UI screenshot to code
python skills/zai-vision/scripts/zai_vision.py ui2code screenshot.png code "implemented with React"
# OCR text extraction
python skills/zai-vision/scripts/zai_vision.py ocr document.png "extract all text"
# Error diagnosis
python skills/zai-vision/scripts/zai_vision.py error crash.png "analyze this error"
# Architecture diagram analysis
python skills/zai-vision/scripts/zai_vision.py diagram architecture.png "explain this architecture"
# UI comparison
python skills/zai-vision/scripts/zai_vision.py diff design.png implementation.png "find differences"
# Video analysis
python skills/zai-vision/scripts/zai_vision.py video demo.mp4 "summarize video content"
```
### Using in AI Coding Tools
#### Claude Code / Gemini CLI
Place the `skills` directory in the project root; AI tools will automatically recognize `SKILL.md` files and understand how to call them:
```
your-project/
├── skills/
│ ├── web-search/
│ │ ├── SKILL.md
│ │ └── scripts/
│ ├── web-reader/
│ │ ├── SKILL.md
│ │ └── scripts/
│ └── ...
└── your-code/
```
#### Custom Agent
Each skill's `SKILL.md` contains structured metadata (YAML frontmatter) for building tool calling systems:
```yaml
---
name: web-search
description: Search the web for real-time information...
---
```
### Project Structure
```
glm-mcp-skills/
├── README.md
├── LICENSE
├── pyproject.toml # Python dependency configuration
└── skills/
├── web-reader/
│ ├── SKILL.md # Skill documentation
│ └── scripts/
│ └── web_reader.py
├── web-search/
│ ├── SKILL.md
│ └── scripts/
│ └── web_search.py
├── zread/
│ ├── SKILL.md
│ └── scripts/
│ └── zread.py
└── zai-vision/
├── SKILL.md
└── scripts/
├── zai_vision.py
└── prompts/ # Preset prompt templates
```
## Contributing
Contributions are welcome! Feel free to:
- 🐛 Report bugs
- 💡 Suggest new skills
- 🔧 Submit pull requests
## License
[MIT License](LICENSE) © 2026 eject
<div align="center">
Made with ❤️ for the AI Coding Community
</
