NPM-agentskills：將AI代理文件與NPM套件打包

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

To see all available qualifiers, see our documentation.

Framework-agnostic skill discovery and export for AI coding agents

License

Uh oh!

There was an error while loading. Please reload this page.

onmax/npm-agentskills

Folders and files

Latest commit

History

Repository files navigation

npm-agentskills

Bundle agent skills with your npm packages to help AI assistants understand your library.

What Are Agent Skills?

Agent Skills provide contextual documentation that AI coding assistants load automatically. When developers use your library, their AI agent reads your skill to deliver accurate guidance on your API, patterns, and best practices.

Skills follow the agentskills open format, supported by Claude Code, GitHub Copilot, Cursor, and other AI coding tools.

This package enables library authors to bundle skills with their npm packages. When users install your package, the skills are discovered and exported to the appropriate location for their AI agent.

Installation

For Library Authors: Bundling Skills

Add Skills to Your Package

Add an agentskills field to your package.json that points to skill directories:

Each skill directory must contain a SKILL.md file with YAML frontmatter:

SKILL.md Format

Your SKILL.md file requires YAML frontmatter with name, description, and license:

Writing Effective Descriptions

The description field is critical. AI agents use it to decide when to activate your skill automatically. Make it specific and actionable.

Good descriptions:

Poor descriptions:

Content Structure

Organize content by task to help AI agents find relevant information quickly:

Code Examples

Include practical, copy-paste examples with file path labels:

Reference Files

Split long documentation into focused reference files:

Reference them from SKILL.md using clear instructions:

Module Configuration

Document all available options with defaults and types:

Auto-imports

List what gets auto-imported to help AI agents understand available utilities:

Integration Patterns

Show how your module works with common tools and patterns:

TypeScript Types

Document exported types for better IDE support:

Multiple Skills Per Package

You can bundle multiple skills for different use cases:

For End Users: Installing Skills

Nuxt Applications (Automatic)

Add the module to your Nuxt config. Skills are automatically discovered from node_modules and exported when you run nuxi prepare or nuxi dev:

The module scans your dependencies for packages with agentskills fields and exports their skills to project-local directories.

Other Frameworks (CLI)

Run the CLI after installing packages that contain skills:

Add this to your postinstall script for automatic exports:

Supported Agents

All paths are project-local, keeping skills scoped to your project rather than polluting global configuration.

Cross-Agent Compatibility

Many agents read from multiple directories for compatibility:

If you target claude, your skills will work with Goose and OpenCode automatically.

Programmatic API

Use the core functions directly in build tools or custom integrations:

CLI Reference

How It Works

For Nuxt, this happens automatically during nuxi prepare. For other frameworks, you run the CLI manually or via npm scripts.

Checklist for Module Authors

Before publishing your package with skills:

Testing Your Skills Locally

Create a test project and verify skills are discovered correctly:

Best Practices

Keep Skills Updated

Update your skills when you:

Focus on Common Tasks

Document the most frequent use cases first:

Use Clear Language

Write for AI agents to parse effectively:

Avoid External Dependencies

Keep all documentation local:

Examples from the Ecosystem

Reference these Nuxt modules for skill patterns:

FAQ

Q: Can I include multiple skills in one package?

Yes. Add multiple entries to the skills array in your package.json:

Q: Should I commit skills to my repository?

Yes. Skills are part of your package's documentation. Users shouldn't need extra packages to get AI assistance with your module.

Q: How do I test skills with my AI agent?

Export skills to your agent's directory and ask questions about your module. Verify the AI provides accurate guidance based on your skill content.

Q: Can skills reference external URLs?

Avoid external links. AI agents work best with local markdown files. Use references/ to split content into focused files instead.

Q: What if my module has many features?

Split documentation across multiple reference files or create separate skills for distinct feature sets. Keep SKILL.md focused on common tasks.

Q: Do skills work with all AI agents?

Skills use the agentskills open format, supported by Claude Code, GitHub Copilot, Cursor, and other AI coding tools. The export paths are standardized across agents.

License

MIT

About

Framework-agnostic skill discovery and export for AI coding agents

Resources

License

Uh oh!

There was an error while loading. Please reload this page.

Stars

Watchers

Forks

Releases

Packages

  0

Languages

Footer

Footer navigation