Repo-to-Skill Converter 🤖

AI-powered Streamlit app to automatically convert GitHub repositories into Claude Skills

Transform any GitHub repository into a comprehensive Claude Skill with just one click! This tool uses Claude AI to analyze repositories and generate professional, well-structured skills following Anthropic's best practices.

Features

✅ AI-Powered Analysis: Intelligently analyzes repository structure, documentation, and code using multiple AI providers

✅ Flexible Authentication: 4 authentication methods supported:

Vertex AI (Gemini 2.5 Flash/Pro, Claude Sonnet 4.5)
Google AI Studio (Gemini models with API key)
Anthropic API (Claude 3/3.5 models)
OpenAI API (GPT-4o, o1 models)

✅ Secure by Design: Production-ready security with input validation, HTTPS-only, and credential protection

✅ Fully Automated: Just provide a GitHub URL - the app handles everything else

✅ Batch Processing: Convert multiple repositories at once

✅ Automatic Validation: Validates skills using Anthropic's official validation script

✅ Skill History: Tracks all generated skills with ability to download or regenerate

✅ One-Click Installation: Automatically install to ~/.claude/skills/ or download as .zip

✅ Modern UI: Beautiful, responsive Streamlit interface with progress tracking

✅ Progressive Disclosure: Generated skills follow Anthropic's pattern with lean SKILL.md and detailed references

Quick Start

Prerequisites

Python 3.10+
Git installed on your system
Choose one authentication method:
- Option 1 (Recommended): Vertex AI
  - Google Cloud Project with Vertex AI API enabled
  - Authenticated with gcloud auth application-default login
  - No API key needed!
- Option 2: Google AI Studio
  - Google API key (Get one here)
- Option 3: Anthropic API
  - Anthropic API key (Get one here)
- Option 4: OpenAI API
  - OpenAI API key (Get one here)

Installation

Clone or navigate to this directory:

cd repo-to-skill

Install dependencies:

pip install -r requirements.txt

Set up authentication:

Option A: Vertex AI

# Install Google Cloud SDK if not already installed
# See: https://cloud.google.com/sdk/docs/install

# Authenticate
gcloud auth application-default login

# Set your project
gcloud config set project YOUR_PROJECT_ID

# Enable required APIs
gcloud services enable aiplatform.googleapis.com

Then create .env file:

cp .env.example .env

Edit .env and set your GCP project:

GCP_PROJECT_ID=your_project_id
GCP_LOCATION=us-central1

Option B: Google AI Studio

cp .env.example .env

Edit .env and add your API key:

GOOGLE_API_KEY=your_api_key_here

Option C: Anthropic API

cp .env.example .env

Edit .env and add your API key:

ANTHROPIC_API_KEY=your_api_key_here

Option D: OpenAI API

cp .env.example .env

Edit .env and add your API key:

OPENAI_API_KEY=your_api_key_here

Run the app:

streamlit run app.py

Open your browser to http://localhost:8501

Usage

Choosing Your AI Model

The app supports multiple AI models via 4 authentication methods:

Via Vertex AI (Recommended):

Claude Sonnet 4.5 - Highest quality (default)
Gemini 2.5 Flash - Fast, cost-effective alternative
Gemini 2.5 Pro - More capable, better for complex repositories

Via Google AI Studio:

Gemini 2.5 Flash - Fast and efficient (default)
Gemini 2.5 Pro - Advanced model
Gemini 2.0 Flash (Experimental) - Latest experimental features

Via Anthropic API:

Claude 3.5 Sonnet - Latest Claude model (default)
Claude 3.5 Haiku - Fast and efficient
Claude 3 Opus - Most capable

Via OpenAI API:

GPT-4o Mini - Efficient and fast (default)
GPT-4o - Latest GPT-4 Optimized
GPT-4 Turbo - High capability
o1-preview - Advanced reasoning model

Single Repository Conversion

Open the app in your browser
Configure authentication in the sidebar:
- Select authentication method: "Vertex AI", "Google AI Studio", "Anthropic API", or "OpenAI API"
- For Vertex AI: Enter your GCP Project ID and location
- For API-based methods: Enter your API key
Select your model from the dropdown
Go to the Generate tab
Paste a GitHub repository URL (e.g., https://github.com/owner/repo)
Click Generate Skill
Wait for the AI to analyze and generate the skill
Choose to install directly or download the .zip file

Batch Conversion

Go to the Batch Generate tab
Enter multiple GitHub URLs (one per line)
Click Generate All Skills
Each repository will be processed sequentially
Download or install each skill individually

View History

Click on History in the sidebar
See all previously generated skills
Re-download skills as needed
View generation statistics

How It Works

The app follows a comprehensive 5-step process:

1. Cloning 🔄

Validates the GitHub URL
Clones the repository to a temporary directory
Fetches repository metadata from GitHub API

2. Analyzing 🔍

Extracts README content
Analyzes file structure
Identifies programming languages
Detects repository type (SDK, framework, library, etc.)
Extracts code samples
Checks for documentation

3. Generating 🤖

Sends repository analysis to Claude API
AI creates comprehensive skill content:
- SKILL.md with workflows and quick starts
- Reference documentation (500-700 lines)
- Code templates with examples
Follows Anthropic's progressive disclosure pattern
Uses imperative voice and objective tone

4. Validating ✅

Validates YAML frontmatter
Checks skill structure
Ensures required files exist
Verifies formatting

5. Packaging 📦

Creates proper skill folder structure
Packages to .zip file
Optionally installs to ~/.claude/skills/
Saves to database for history tracking

Generated Skill Structure

Each generated skill follows this professional structure:

skill-name/
├── SKILL.md                      # Main skill file with workflows
├── references/
│   ├── implementation-guide.md   # Detailed implementation docs
│   ├── api-reference.md          # API or concept reference
│   └── advanced-patterns.md      # Advanced usage patterns
└── assets/templates/
    ├── basic-example.py          # Basic working example
    ├── advanced-example.py       # Advanced use case
    └── integration-example.py    # Integration pattern

SKILL.md Contents

YAML Frontmatter: Name and description
Overview: When to use this skill
Quick Start: Installation and basic example
Core Workflows: 3-5 step-by-step workflows
Key Concepts: Main concepts explained
Reference Documentation: Links to detailed docs
Templates: Links to code examples
Best Practices: Dos and don'ts
Troubleshooting: Common issues

Configuration

Authentication Setup

Option 1: Vertex AI (Recommended)

Advantages:

No API key management needed
Access to both Gemini and Claude models
Lower costs with Gemini Flash
Enterprise-grade security with ADC

Setup:

Install Google Cloud SDK:

# macOS
brew install google-cloud-sdk

# Or download from: https://cloud.google.com/sdk/docs/install

Authenticate:

# Login and create application default credentials
gcloud auth application-default login

# Set your project
gcloud config set project YOUR_PROJECT_ID

Enable APIs:

# Enable Vertex AI
gcloud services enable aiplatform.googleapis.com

# Enable Gen AI (for Gemini/Claude models)
gcloud services enable generativelanguage.googleapis.com

Set environment variables:

Create .env file:

GCP_PROJECT_ID=your_project_id
GCP_LOCATION=us-central1  # Or your preferred region
GITHUB_TOKEN=your_github_token  # Optional

Supported Models:

claude-sonnet-4-5@20250929 - Claude via Vertex AI (default)
gemini-2.5-flash - Fast, cost-effective
gemini-2.5-pro - High capability

Option 2: Google AI Studio

Advantages:

Simple setup with API key
Direct access to latest Gemini models
No Google Cloud account needed
Free tier available

Setup:

Get API key from aistudio.google.com/apikey
Create .env file:

GOOGLE_API_KEY=your_api_key_here
GITHUB_TOKEN=your_github_token  # Optional

Supported Models:

gemini-2.5-flash - Fast and efficient (default)
gemini-2.5-pro - Advanced model
gemini-2.0-flash-exp - Experimental features

Option 3: Anthropic API

Advantages:

Simple setup
Direct access to Claude models
Latest Claude features

Setup:

Get API key from console.anthropic.com
Create .env file:

ANTHROPIC_API_KEY=your_api_key_here
GITHUB_TOKEN=your_github_token  # Optional

Supported Models:

claude-3-5-sonnet-20241022 - Latest Sonnet (default)
claude-3-5-haiku-20241022 - Fast and efficient
claude-3-opus-20240229 - Most capable

Option 4: OpenAI API

Advantages:

Access to GPT-4o and o1 models
Simple API key setup
Advanced reasoning with o1

Setup:

Get API key from platform.openai.com/api-keys
Create .env file:

OPENAI_API_KEY=your_api_key_here
GITHUB_TOKEN=your_github_token  # Optional

Supported Models:

gpt-4o-mini - Efficient and fast (default)
gpt-4o - Latest GPT-4 Optimized
gpt-4-turbo - High capability
o1-preview - Advanced reasoning

Environment Variables Reference

Create a .env file with:

# === Vertex AI Configuration (Option 1) ===
GCP_PROJECT_ID=your_project_id
GCP_LOCATION=us-central1

# === Google AI Studio Configuration (Option 2) ===
GOOGLE_API_KEY=your_google_api_key

# === Anthropic API Configuration (Option 3) ===
ANTHROPIC_API_KEY=your_anthropic_api_key

# === OpenAI API Configuration (Option 4) ===
OPENAI_API_KEY=your_openai_api_key

# === Optional ===
# GitHub Personal Access Token (for higher rate limits)
GITHUB_TOKEN=your_github_token_here

Settings in config.py

You can customize various settings in config.py:

DEFAULT_VERTEX_MODEL: Default AI model (default: claude-sonnet-4-5@20250929)
MAX_TOKENS: Max tokens for generation (default: 32000)
CLAUDE_SKILLS_DIR: Installation directory (default: ~/.claude/skills)
MAX_BATCH_SIZE: Maximum repositories in batch mode (default: 5)
OUTPUT_DIR: Where generated skills are saved
Authentication methods and available models

Project Structure

repo-to-skill/
├── app.py                    # Main Streamlit application
├── config.py                 # Configuration and constants
├── requirements.txt          # Python dependencies
├── .env.example             # Environment variables template
├── README.md                # This file
├── src/
│   ├── __init__.py
│   ├── repo_analyzer.py     # Repository cloning and analysis
│   ├── ai_generator.py      # Claude API integration
│   ├── skill_builder.py     # Skill structure creation
│   ├── validator.py         # Skill validation
│   ├── packager.py          # Packaging and installation
│   └── database.py          # SQLite database operations
├── templates/
│   ├── skill_prompt.txt     # AI generation prompt
│   └── custom_styles.css    # UI styling
├── data/
│   ├── skills_history.db    # SQLite database (auto-created)
│   └── temp/                # Temporary repository clones
└── output/
    └── generated/           # Generated .zip files and skills

Examples

Example 1: Convert Python SDK

URL: https://github.com/anthropics/anthropic-sdk-python

Generated Skill:
- Name: anthropic-python-sdk
- 3 core workflows (Installation, Basic Usage, Advanced Patterns)
- 2 reference docs (API Reference, Authentication Guide)
- 3 templates (Basic client, Streaming, Function calling)

Example 2: Convert Framework

URL: https://github.com/google/adk-python

Generated Skill:
- Name: adk-python-framework
- 4 core workflows (Build Agent, Multi-Agent, Custom Tools, Deploy)
- 3 reference docs (Core Concepts, Quick Start, Deployment)
- 3 templates (Basic agent, Multi-agent, Custom tools)

Troubleshooting

App won't start

Issue: ModuleNotFoundError

Solution: Install dependencies

pip install -r requirements.txt

API Key errors

Issue: "Invalid API key"

Solutions:

Check your API key in .env file
Verify key is active at https://console.anthropic.com/
Enter key directly in the sidebar

Repository cloning fails

Issue: "Failed to clone repository"

Solutions:

Ensure Git is installed
Check repository URL is correct
Verify repository is public
Check network connection

Validation fails

Issue: "Validation errors"

Solutions:

This is usually not critical - skill may still work
Check generated SKILL.md for proper YAML frontmatter
Ensure references/ and templates/ have content

Advanced Usage

Custom Validation Script

By default, the app uses Anthropic's validation script from the parent directory. To use a custom validator:

Update VALIDATION_SCRIPT path in config.py
Ensure script accepts skill directory as argument

Custom Skill Prompt

Modify templates/skill_prompt.txt to change how AI generates skills:

Adjust structure requirements
Change reference documentation format
Modify code template requirements

Database Management

The SQLite database stores all generation history. Access it directly:

from src.database import SkillDatabase

db = SkillDatabase()
skills = db.get_all_skills()
stats = db.get_stats()

Best Practices

For Best Results

✅ Use well-documented repositories: The better the README, the better the generated skill

✅ Convert established projects: Mature repositories with examples generate better skills

✅ Review AI output: Always review generated content before using

✅ Provide GitHub token: Avoid rate limits with a personal access token

✅ Start with single generation: Test with one repo before batch processing

What Works Best

SDKs and libraries: Excellent results with clear APIs
Frameworks: Good results with comprehensive docs
Example repositories: Great for sample catalogs
Well-documented projects: Better than sparsely documented ones

Limitations

Public repositories only: Cannot access private repos
Repository size: Very large repos (>1000 files) may take longer
API costs: Each skill generation uses Claude API tokens
Language support: Best results with Python, JavaScript, TypeScript, Go
Documentation dependency: Quality depends on repository documentation

FAQ

Q: Which authentication method should I use?

A: Recommended order:

Vertex AI - Best if you have GCP access (no API key management, access to multiple models)
Google AI Studio - Simple API key setup, free tier available for Gemini models
Anthropic API - Direct access to latest Claude models
OpenAI API - If you need GPT-4o or advanced reasoning with o1

Q: Which model should I choose?

A: For best quality: Claude Sonnet 4.5 (Vertex AI) or Claude 3.5 Sonnet (Anthropic API)

For cost-effectiveness: Gemini 2.5 Flash (Vertex AI or Google AI Studio) or GPT-4o Mini (OpenAI)

For complex repositories: Gemini 2.5 Pro, Claude 3 Opus, or GPT-4o

Q: How much does it cost to generate a skill?

A: Vertex AI:

Gemini 2.5 Flash: ~$0.01-0.05 per skill
Gemini 2.5 Pro: ~$0.05-0.20 per skill
Claude Sonnet 4.5: ~$0.10-0.50 per skill

Google AI Studio:

Gemini models: Free tier available, then similar to Vertex AI pricing

Anthropic API:

Claude 3.5 Sonnet: ~$0.10-0.50 per skill
Claude 3.5 Haiku: ~$0.05-0.15 per skill
Claude 3 Opus: ~$0.50-1.50 per skill

OpenAI API:

GPT-4o Mini: ~$0.02-0.10 per skill
GPT-4o: ~$0.20-0.80 per skill
o1-preview: ~$0.50-2.00 per skill

(Costs vary based on repository size and documentation)

Q: Can I customize the generated skills?

A: Yes! Generated skills are in output/generated/ - edit before packaging or installing.

Q: Will this work with private repositories?

A: Not currently. Only public GitHub repositories are supported.

Q: Can I regenerate a skill?

A: Yes, just enter the same URL again. The app will create a new version.

Q: Where are skills installed?

A: ~/.claude/skills/ directory (configurable in config.py)

Security

This application implements production-ready security controls:

✅ Input Validation: All GitHub URLs validated with regex, length limits, and HTTPS-only enforcement

✅ Path Traversal Prevention: Sanitized paths with containment verification

✅ Resource Limits: 500MB repository limit, 1MB README limit, file size restrictions

✅ Command Injection Prevention: Safe GitPython API usage, no shell execution

✅ SQL Injection Prevention: Parameterized queries throughout

✅ Credential Protection: Masked API key inputs, no logging of sensitive data, environment variable support

✅ Secure Defaults: HTTPS-only, strict validation, automatic cleanup

For detailed security information, see SECURITY.md.

Contributing

Contributions welcome! Areas for improvement:

Support for private repositories
GitLab/Bitbucket support
More customizable templates
Skill editing before generation
Better error handling
Performance optimizations

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Anthropic for Claude AI and Skills framework
Streamlit for the amazing web framework
GitHub for repository hosting and API

Support

For issues or questions:

Check the Troubleshooting section
Review generated logs in the app
Check Streamlit console output for errors

Built with ❤️ using Claude AI and Streamlit

Version 1.0.0

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
output/generated		output/generated
src		src
templates		templates
.env.example		.env.example
.gitignore		.gitignore
AUTHENTICATION.md		AUTHENTICATION.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
VERTEX_AI_SETUP.md		VERTEX_AI_SETUP.md
app.py		app.py
config.py		config.py
requirements.txt		requirements.txt
run.bat		run.bat
run.sh		run.sh

License

eddiehaug/repo-to-claude-skill

Folders and files

Latest commit

History

Repository files navigation

Repo-to-Skill Converter 🤖

Features

Quick Start

Prerequisites

Installation

Usage

Choosing Your AI Model

Single Repository Conversion

Batch Conversion

View History

How It Works

1. Cloning 🔄

2. Analyzing 🔍

3. Generating 🤖

4. Validating ✅

5. Packaging 📦

Generated Skill Structure

SKILL.md Contents

Configuration

Authentication Setup

Option 1: Vertex AI (Recommended)

Option 2: Google AI Studio

Option 3: Anthropic API

Option 4: OpenAI API

Environment Variables Reference

Settings in config.py

Project Structure

Examples

Example 1: Convert Python SDK

Example 2: Convert Framework

Troubleshooting

App won't start

API Key errors

Repository cloning fails

Validation fails

Advanced Usage

Custom Validation Script

Custom Skill Prompt

Database Management

Best Practices

For Best Results

What Works Best

Limitations

FAQ

Security

Contributing

License

Acknowledgments

Support

About

Resources

License

Security policy

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Contributors 2

Languages

Packages