OpenSkill CLI

Create and manage Claude skills with AI-powered content generation

Installation • Quick Start • Commands • Configuration • Contributing

---

What is OpenSkill?

OpenSkill is a command-line tool that simplifies creating and managing Claude skills. It uses AI to automatically generate comprehensive skill descriptions and rules from simple prompts.

Features

• Multi-Provider AI - Choose from Groq, OpenAI, Anthropic, or Ollama (local)
• SKILL.md Format - Skills stored as Markdown with YAML frontmatter for Claude's native skill discovery
• Version History - Track changes with automatic versioning and rollback support
• Skill Composition - Extend and combine skills with extends and includes
• Validation - Validate skill structure before deploying
• Fast - Built in Go for maximum performance

Supported AI Providers

Provider	Model (Default)	API Key
Groq	llama-3.3-70b-versatile	console.groq.com
OpenAI	gpt-4o-mini	platform.openai.com
Anthropic	claude-3-5-sonnet-20241022	console.anthropic.com
Ollama	llama3.2	No API key (runs locally)

Installation

Quick Install (Recommended)

curl -fsSL openskill.online/api/install | bash

Build from Source

Requires Go 1.21+

git clone https://github.com/rakshit-gen/openskill.git
cd openskill && make build && sudo make install

Verify Installation

openskill --help

Configuration

Set Up Your AI Provider

OpenSkill supports multiple AI providers. Set your preferred provider and API key:

# Set provider (default: groq)
openskill config set provider groq    # or: openai, anthropic, ollama

# Set API key for your provider
openskill config set api-key

Provider-Specific Setup

Groq (Default - Free & Fast)

openskill config set provider groq
openskill config set groq-api-key YOUR_KEY

OpenAI

openskill config set provider openai
openskill config set openai-api-key YOUR_KEY

Anthropic

openskill config set provider anthropic
openskill config set anthropic-api-key YOUR_KEY

Ollama (Local - No API Key)

# Make sure Ollama is running: ollama serve
openskill config set provider ollama
openskill config set ollama-model llama3.2

View Configuration

openskill config list

Environment Variables

Environment variables take precedence over config file:

Variable	Description
OPENSKILL_PROVIDER	Active provider (groq, openai, anthropic, ollama)
GROQ_API_KEY	Groq API key
OPENAI_API_KEY	OpenAI API key
ANTHROPIC_API_KEY	Anthropic API key
OPENSKILL_MODEL	Override model for any provider
OLLAMA_HOST	Custom Ollama endpoint

Quick Start

Initialize OpenSkill

openskill init

Create your first skill

openskill add "code-review" -d "Reviews code for best practices"

The AI will generate a comprehensive skill:

Generating skill with Groq...

✓ Added skill: code-review
  Description: Comprehensive code review focusing on security, performance, and maintainability
  Rules:
    1. Check for security vulnerabilities (XSS, SQL injection, etc.)
    2. Verify proper error handling and edge cases
    3. Ensure code follows project conventions
    4. Review test coverage and quality

List all skills

openskill list

View skill details

openskill show "code-review"

Commands

Command	Description
openskill init	Initialize OpenSkill in your project
openskill add <name> -d <description>	Create a new skill with AI generation
openskill add <name> -d <desc> --manual -r <rule>	Create skill manually with custom rules
openskill list	List all skills
openskill show <name>	Show detailed skill information
openskill edit <name> -d <description>	Update skill description
openskill edit <name> -r <rule1> -r <rule2>	Replace skill rules
openskill remove <name>	Delete a skill
openskill validate <name>	Validate skill structure
openskill history <name>	Show version history
openskill rollback <name> <version>	Restore a previous version
openskill config set <key> [value]	Set configuration
openskill config get <key>	Get configuration value
openskill config list	List all configuration

Flags

Flag	Description
-d, --desc	Skill description (required for add)
-r, --rule	Add a rule (can be repeated, manual mode only)
--manual	Skip AI generation, use provided values

Skill Format

Skills are stored as Markdown files with YAML frontmatter in .claude/skills/<name>/SKILL.md:

.claude/
└── skills/
    ├── code-review/
    │   └── SKILL.md
    └── bug-finder/
        └── SKILL.md

SKILL.md Structure

---
name: code-review
description: Comprehensive code review focusing on security and maintainability
---

# code-review

Comprehensive code review focusing on security, performance,
and maintainability best practices.

## Rules

- Check for security vulnerabilities (XSS, SQL injection, etc.)
- Verify proper error handling and edge cases
- Ensure code follows project conventions
- Review test coverage and quality

Skill Composition

Extend skills with extends:

---
name: security-review
description: Security-focused code review
extends: code-review
---

# security-review

## Rules

- Focus on OWASP Top 10 vulnerabilities
- Check authentication and authorization flows

Combine skills with includes:

---
name: full-review
description: Comprehensive review combining multiple aspects
includes:
  - code-review
  - security-review
  - performance-review
---

# full-review

## Rules

- Provide a summary score for each review area

Project Structure

openskill/
├── cmd/
│   └── openskill/
│       ├── main.go           # Entry point
│       └── commands/         # CLI commands
│           ├── init.go       # Initialize project
│           ├── add.go        # Add skills
│           ├── list.go       # List skills
│           ├── show.go       # Show skill details
│           ├── edit.go       # Edit skills
│           ├── remove.go     # Remove skills
│           ├── validate.go   # Validate skills
│           ├── history.go    # Version history
│           ├── rollback.go   # Rollback versions
│           └── config.go     # Configuration
├── pkg/
│   ├── core/
│   │   └── skill.go          # Skill data structure
│   ├── skills/
│   │   └── manager.go        # Skill file management
│   ├── llm/
│   │   ├── provider.go       # Provider interface
│   │   ├── generator.go      # AI generation
│   │   ├── groq.go           # Groq client
│   │   ├── openai.go         # OpenAI client
│   │   ├── anthropic.go      # Anthropic client
│   │   └── ollama.go         # Ollama client
│   └── config/
│       └── config.go         # Configuration management
├── Makefile
├── go.mod
└── go.sum

Contributing

We welcome contributions! Please see our Contributing Guide for details.

Development Setup

1. Fork and clone the repository
2. Install Go 1.21+
3. Run make deps to install dependencies
4. Run make build to build
5. Run make test to run tests

Quick Commands

make build      # Build binary
make install    # Install to /usr/local/bin
make test       # Run tests
make clean      # Clean build artifacts
make deps       # Tidy dependencies
make lint       # Run linter
make release    # Build for all platforms

License

MIT License - see LICENSE for details.

Acknowledgments

• Groq for fast LLM inference
• OpenAI for GPT models
• Anthropic for Claude models
• Ollama for local LLM support
• Cobra for CLI framework
• Claude for inspiration

---

Made with ❤️ by Rakshit-gen