Roo Skills

Bring Anthropic's Agent Skills to Roo Code

An NPM package and template that enables Roo Code users to leverage Anthropic's Agent Skills system - the same powerful capability available in Claude Code.

Features

• 🚀 One-Command Setup - Install skills in any project with npx @kastalien-research/rooskills init
• 🤖 Automated Generation - Create new skills from any documentation URL
• 📚 Pre-built Skills - Ready-to-use skills for development, AI automation, and documentation
• 🔧 CLI Tools - Simple commands for initialization and skill generation
• ⚡ Batch Automation - Run multiple research iterations in parallel with Sequential Feynman Batch Executor
• 🔬 Deep Research - Automated deep-learning workflows with validation and artifact collection

Quick Start

Installation (Recommended)

Install skills in your Roo Code project with a single command:

npx @kastalien-research/rooskills init

This will:

• Copy all pre-built skills to .roo/skills/
• Add .roomodes configuration for Roo Code
• Create .env.example with required API keys
• Optionally install Python dependencies

Generate a New Skill

Create a custom skill from any documentation URL:

npx @kastalien-research/rooskills generate https://docs.example.com skill-name --max-urls 20

List Available Skills

npx @kastalien-research/rooskills list

What are Agent Skills?

Agent Skills are modular instruction sets that enhance Claude's capabilities for specific tasks. Originally developed for Claude Code, this package brings that same functionality to Roo Code by:

1. Converting Agent Skills into Roo Code custom modes via .roomodes
2. Organizing skills in .roo/skills/ for clean project structure
3. Providing automated tools to generate new skills from documentation

Pre-built Skills

All skills are located in .roo/skills/ and registered as custom modes in .roomodes:

Development & Technical

• mcp-builder - Create high-quality MCP (Model Context Protocol) servers for integrating external services
• model-enhancement-mcp - Advanced MCP server examples including analogical reasoning and sequential thinking
• document-skills - Comprehensive document creation and manipulation (DOCX, PDF, PPTX, XLSX)

AI & Automation

• lindy-expert - Expertise in Lindy AI agent creation and management platform for business process automation

Development Tools

• create-llmstxt-py - Python tool for generating llms.txt documentation from websites
• skill-creator - Meta-skill for creating new effective skills with best practices
• template-skill - Basic template for starting new skills

Documentation

• architecture - Architecture documentation for the skill generation system

Automation Scripts

Sequential Feynman Batch Executor

Run multiple deep research iterations automatically with the Sequential Feynman workflow:

# Basic usage - single iteration
./scripts/feynman-batch.sh "distributed tracing systems"

# Multiple iterations for comparative analysis
./scripts/feynman-batch.sh "React Server Components" 3

# Parallel execution (2 concurrent iterations)
./scripts/feynman-batch.sh "MCP architecture" 5 --parallel 2

# Background execution
nohup ./scripts/feynman-batch.sh "WebAssembly optimization" 3 > batch.log 2>&1 &

Features:

• ⚡ Parallel or sequential execution modes
• 🔄 Automatic retry logic (3 attempts per iteration)
• 📊 Real-time progress tracking via STATUS.json
• 📝 Comprehensive summary reports
• 🎯 Artifact collection (notebooks, skills, outputs)
• ⏱️ Configurable timeouts and concurrency

Output Structure:

feynman-batch-outputs/
  [run-id]/
    iteration-1/
      output.json
      artifacts/
        feynman-*.ipynb
        skill-name/
    STATUS.json
    REPORT.md

See scripts/README-feynman-batch.md for complete documentation.

Advanced Usage

Environment Variables

After initialization, create a .env file in your project root:

# Firecrawl API Key (get from https://firecrawl.dev)
FIRECRAWL_API_KEY=your_key_here

# OpenAI API Key (get from https://platform.openai.com)
OPENAI_API_KEY=your_key_here

# Anthropic API Key (get from https://console.anthropic.com)
ANTHROPIC_API_KEY=your_key_here

# Optional: Exa API Key for enhanced research
EXA_API_KEY=your_key_here

Skill Generation Process

The skill generator runs through 5 automated phases:

1. Knowledge Extraction - Scrapes and summarizes documentation using Firecrawl + OpenAI
2. Ecosystem Research - Analyzes positioning and best practices using Claude + Exa
3. Skill Synthesis - Generates structured SKILL.md with comprehensive guidance
4. File Writing - Creates all skill files and reference documentation
5. Mode Registration - Registers the skill as a mode in .roomodes

Generated Skill Structure

Each generated skill includes:

.roo/skills/skill-name/
├── SKILL.md                          # Main skill file with YAML frontmatter
├── LICENSE.txt                       # License information
└── references/
    ├── api_documentation.md          # Complete API/documentation content
    └── documentation_index.md        # Curated index of key resources

Manual Installation (Alternative)

If you prefer not to use the NPM package:

1. Fork or Clone this repository to your Roo Code workspace
2. Copy .env.example to .env and add your API keys
3. Import the .roomodes configuration into your Roo Code settings
4. Skills in .roo/skills/ will be available as custom modes

For Claude Code Users

While this package is designed for Roo Code, the skills themselves are compatible with Claude Code's native Agent Skills system:

1. Skills can be used directly from Anthropic's official skills repository
2. Or you can copy individual skills from this template's .roo/skills/ directory

Using Skills in Roo Code

Skills are automatically loaded from .roomodes and can be activated by:

1. Mentioning the skill name in your conversation
2. Switching to the skill's mode using the mode selector
3. Referencing skills in task delegation

CLI Reference

rooskills init

Initialize Roo Skills in your project.

Options:

• -d, --dir <directory> - Target directory (default: current directory)
• --no-install - Skip Python dependency installation
• --skills <skills...> - Install specific skills only

Examples:

rooskills init
rooskills init -d ./my-project
rooskills init --no-install

rooskills generate <url> [skill-name]

Generate a new skill from documentation URL.

Arguments:

• url - Documentation URL to process (required)
• skill-name - Skill identifier in kebab-case (optional)

Options:

• --max-urls <number> - Maximum URLs to process (default: 20)
• --output-dir <directory> - Custom output directory
• --verbose - Enable verbose logging

Examples:

rooskills generate https://fastapi.tiangolo.com
rooskills generate https://docs.stripe.com stripe-expert --max-urls 30

rooskills list

List all available pre-built skills.

Validation

Validate skill integration:

python .roo/skills/scripts/validate_mode_skill_integration.py

This checks:

• SKILL.md file exists and is properly formatted
• YAML frontmatter is valid
• No duplicate slugs in .roomodes
• File sizes are within limits (<500 lines)
• Skills are properly registered

Project Structure

rooskills/
├── .env                              # Environment variables (git-ignored)
├── .roomodes                         # Mode configuration file
├── README.md                         # This file
├── agent_skills_spec.md              # Skills specification
└── .roo/
    └── skills/                       # All skills directory
        ├── architecture/             # Architecture documentation
        ├── create-llmstxt-py/        # Python tool for llms.txt generation
        ├── document-skills/          # Document creation (DOCX, PDF, PPTX, XLSX)
        ├── lindy-expert/             # Lindy AI automation platform skill
        ├── mcp-builder/              # MCP server creation skill
        ├── model-enhancement-mcp/    # Advanced MCP examples
        ├── skill-creator/            # Meta-skill for creating skills
        ├── template-skill/           # Basic skill template
        └── scripts/                  # Skill generation scripts
            ├── agent-skill-generator/
            │   ├── config.py         # Configuration management
            │   ├── orchestrator.py   # Main CLI entry point
            │   ├── llms_generator.py # Knowledge extraction
            │   ├── ecosystem_researcher.py
            │   ├── skill_creator.py  # Skill synthesis
            │   └── mode_configurator.py
            ├── commands/
            │   └── generate-skill.sh # Shell wrapper
            └── validate_mode_skill_integration.py

Architecture

The skill generator follows a modular, pipeline-based architecture:

• Configuration - Environment-based settings (never hardcoded secrets)
• Knowledge Extraction - Firecrawl + OpenAI for documentation processing
• Research - Claude + Exa for ecosystem analysis
• Synthesis - Structured skill creation with best practices
• Validation - Comprehensive integration checks

See architecture/ for detailed documentation.

Best Practices

When Creating Skills

• Keep SKILL.md under 500 lines - move detailed content to references/
• Use clear, descriptive frontmatter (name, description)
• Provide specific examples and guidelines
• Include integration patterns and common pitfalls
• Never hardcode secrets or environment-specific values

When Generating Skills

• Start with smaller --max-urls values (10-15) for faster results
• Review generated content before committing
• Customize the SKILL.md to add organization-specific knowledge
• Test the skill thoroughly before deploying

Contributing

Skills in this repository use modular architecture principles:

• Files should be < 500 lines
• Use environment variables for configuration
• Follow clean architecture patterns
• Document all functionality clearly

CI/CD & Security

This package includes comprehensive CI/CD workflows with automated security checks:

Automated Workflows

• CI/CD Pipeline - Runs on every push and PR

  • Security validation (no exposed API keys, .env files, or secrets)
  • Package integrity checks
  • CLI functionality tests
  • Automated package validation

• PR Validation - Validates pull requests

  • Checks for sensitive file changes
  • Package size monitoring
  • Version change detection

• Automated Publishing - Publishes to npm

  • Pre-publish security scans
  • npm provenance support
  • Automatic git tagging
  • Package artifact archival

Security Features

✅ No secrets in repository - All API keys are gitignored
✅ Git history scanning - Prevents accidental .env commits
✅ Package validation - Ensures no sensitive files in npm package
✅ Python cache exclusion - No __pycache__ or .pyc files
✅ IDE artifact filtering - Excludes .claude/, .refactoring-game/

See .github/workflows/README.md for detailed CI/CD documentation.

License

See individual skill directories for licensing information. Most skills use MIT License unless otherwise specified.

Resources & References

Anthropic's Agent Skills

• Official Agent Skills Repository - Original skills for Claude Code
• What are Skills? - Understanding the skills system
• Equipping Agents for the Real World - Engineering blog post
• Creating Custom Skills - Official guide

This Template

• Architecture Documentation - Skill generation system design
• Validation Scripts - Testing and validation tools
• Skill Creator Skill - Meta-skill for creating new skills

Compatibility

Agent Skills Format

This template uses Anthropic's Agent Skills format (YAML frontmatter + Markdown instructions), ensuring skills are:

• Portable - Work in both Roo Code and Claude Code (with appropriate configuration)
• Standard - Follow Anthropic's established patterns and best practices
• Maintainable - Compatible with future skill format updates

Roo Code Integration

Skills integrate with Roo Code through:

• Custom Modes - Each skill becomes a mode in .roomodes
• Skill References - Skills are loaded dynamically via skill_ref paths
• Override Strategy - Skill content overrides default mode instructions

Support

For issues, questions, or contributions:

1. Check existing documentation
2. Review validation scripts
3. Consult the skill creator skill
4. Reference Anthropic's official skills documentation

---

Note: This is a template repository that bridges Anthropic's Agent Skills system with Roo Code. Skills are compatible with the official Agent Skills format and can be shared between Claude Code and Roo Code environments.