Agent Skills

The Agent Skills specification is a standard for providing structured documentation to AI coding assistants. It helps AI tools understand your project's APIs, conventions, and best practices through a well-organized directory of markdown files.

Web Awesome publishes an Agent Skill that provides AI tools with comprehensive information about our components, including their APIs, usage patterns, theming options, and more.

This feature is experimental! The Agent Skills format and its contents may change as we refine the output based on feedback and evolving AI capabilities.

Why Use It?

When working with AI coding assistants like Claude Code, Cursor, or other tools that support Agent Skills, you can reference the Web Awesome skill to give the AI deep context about our components. This leads to more accurate code suggestions, better understanding of component APIs, and fewer hallucinations when generating Web Awesome code.

Unlike a single file, the Agent Skill uses progressive disclosure — the AI loads only the documentation it needs for the current task, making it more efficient for complex projects.

Accessing the Skill

The Agent Skill is available in every Web Awesome build at:

/dist/skills/webawesome/
/dist-cdn/skills/webawesome/

You can also find it in your node_modules directory if you've installed Web Awesome via npm:

node_modules/@awesome.me/webawesome/dist/skills/webawesome/

The skill directory contains:

webawesome/
├── SKILL.md              # Main skill file with overview and quick reference
└── references/
    ├── components/       # Individual documentation for each component
    ├── frameworks/       # React, Vue, Angular, Svelte guides
    ├── utilities/        # Layout and styling utilities
    ├── installation.md   # Installation guide
    ├── usage.md          # Usage patterns
    ├── form-controls.md  # Form integration
    ├── themes.md         # Themes and color palettes
    └── ...

How to Use It

How you reference the skill depends on which AI tool you're using.

Claude Code

If you're using Claude Code, you can install the Web Awesome skill locally using the skills CLI:

# Install the skill
npx skills add ./node_modules/@awesome.me/webawesome/dist/skills/webawesome

# Uninstall the skill
npx skills remove webawesome

Once installed, the skill will be available to Claude Code automatically. The skill is installed as a symlink, so it will automatically stay up to date when you update Web Awesome via npm.

You can also reference the skill directory manually:

@node_modules/@awesome.me/webawesome/dist/skills/webawesome/

Or ask Claude to read specific reference files:

@node_modules/@awesome.me/webawesome/dist/skills/webawesome/references/components/button.md

Claude Code can also discover and use the skill automatically if it's in your project.

Cursor

In Cursor, you can add the skill directory to your project's documentation sources via Cursor Settings > Features > Docs. Point it to the dist/skills/webawesome/ directory in your node_modules folder.

Other AI Tools

Most AI coding assistants that support the Agent Skills specification can use this skill. Check your tool's documentation for how to add custom skills or documentation sources.

What's Included

The Agent Skill contains:

Agent Skills vs llms.txt

Web Awesome provides both an llms.txt file and an Agent Skill. Here's when to use each:

Feature Agent Skill llms.txt
Format Directory with multiple markdown files Single text file
Best for Deep integration, complex tasks Quick context, simple queries
Context usage Progressive disclosure (loads as needed) Loads everything at once
Supported by Tools supporting agentskills.io spec Most AI tools

If your AI tool supports Agent Skills, we recommend using the skill for better context efficiency. Otherwise, the llms.txt file is a great fallback.

Feedback

Since this is experimental, we'd love to hear how it works for you! If you find issues with the generated content or have suggestions for improvement, please open an issue on GitHub.

    No results
    Navigate Enter Select Esc Close