Learn to use Skills
Everything you need to know about finding, installing, creating, and sharing skills for your AI coding agents.
What Are Skills?
Skills are reusable capabilities for AI coding agents. Unlike static documentation or simple prompts, skills provide procedural knowledge—they teach your AI assistant how to do things, not just facts about things.
Think of skills like plugins for your AI
Just as browser extensions add new features to Chrome, skills add new capabilities to AI agents. They come with instructions, scripts, and configurations that enable your agent to interact with external services and automate complex workflows.
Example Skills
- GitHub integration — Create issues, open PRs, manage repositories
- Weather APIs — Fetch forecasts and conditions from any location
- Database management — Query, backup, and migrate databases
- Deployment automation — Deploy to Vercel, AWS, Railway, and more
- Notification services — Send Slack messages, emails, or SMS alerts
Cross-Agent Compatibility
Skills are designed to work across multiple AI coding agents:
Anatomy of a Skill
At minimum, every skill contains a SKILL.md file. More complex skills include additional scripts and configurations:
my-skill/
├── SKILL.md # Required: skill definition & docs
├── scripts/
│ ├── setup.sh # Optional: installation script
│ └── run.py # Optional: execution script
├── config/
│ └── settings.json # Optional: configuration
└── examples/
└── usage.md # Optional: usage examplesInstalling Skills for Claude Desktop
There are two ways to install skills in Claude Desktop: using the built-in GUI or manually placing files. Once installed, skills become available as slash commands.
Option AUsing the GUI (Recommended)
- 1Open Claude Desktop and go to Settings (gear icon)
- 2Click on Capabilities
- 3Scroll down to the Skills section
- 4Click the + button to add a new skill
- 5Upload the
SKILL.mdfile (and any additional files)
Option BManual Installation
For power users, you can manually place skill files in Claude's commands directory.
Global Skills (all projects)
# Create the directory and add your skill
$ mkdir -p ~/.claude/commands
$ cp weather.md ~/.claude/commands/
Project-Specific Skills
# In your project root
$ mkdir -p .claude/commands
$ cp deploy.md .claude/commands/
Using Your Installed Skill
Once installed, type the slash command (the filename without .md) to invoke the skill:
Example usage:
/weather PittsburghClaude will read the skill instructions and execute the task accordingly.
Directory Structure
~/.claude/
└── commands/ # Global skills (all projects)
├── weather.md → /weather
├── deploy.md → /deploy
└── git-commit.md → /git-commit
your-project/
└── .claude/
└── commands/ # Project-specific skills
└── test.md → /testSubmitting Skills
Ready to share your skill with the community? There are two ways to submit skills to the registry.
Option A: File Upload
- 1Go to the Submit page
- 2Drag and drop your skill folder (or select files)
- 3Include a
SKILL.mdwith proper frontmatter - 4Additional files (scripts, configs) are bundled with your submission
- 5Submissions go through review before appearing in the directory
Option B: GitHub Link
Coming SoonSoon you'll be able to link directly to a GitHub repository:
- Direct linking to your public GitHub repo
- Automatic syncing when your repo updates
- Version tracking tied to git tags
SKILL.md Format
Every skill requires a SKILL.md file. This file uses YAML frontmatter for metadata and Markdown for documentation.
Required Fields
| Field | Type | Description |
|---|---|---|
name | string | Human-readable name of the skill |
description | string | Brief description of what it does |
version | string | Semantic version (e.g., "1.0.0") |
Complete Example
--- name: Weather API Skill description: Fetch weather data from OpenWeatherMap API version: 1.2.0 author: weatherapi category: api agents: - claude-code - cursor - codex --- # Weather API Skill This skill enables your AI agent to fetch weather data. ## Setup 1. Get an API key from OpenWeatherMap 2. Set the environment variable: export OPENWEATHER_API_KEY="your-api-key" ## Usage Ask your agent: - "What's the weather in Pittsburgh?" - "Will it rain tomorrow in New York?"
Best Practices
Follow these guidelines to create skills that are useful, maintainable, and work well across different AI agents.
Stay Focused
Keep each skill focused on one capability. A skill that does one thing well is more reusable.
Clear Examples
Include usage examples in your SKILL.md. Show what prompts users can give.
Document Dependencies
Clearly list any environment variables, API keys, or system dependencies required.
Test Across Agents
If possible, test your skill with multiple AI agents. Note any agent-specific quirks.
Handle Errors Gracefully
Provide helpful error messages when things go wrong. Guide users to the solution.
Version Semantically
Use semantic versioning (MAJOR.MINOR.PATCH). Bump major for breaking changes.
Ready to create your first skill?
Browse existing skills for inspiration, then submit your own to share with the community.