Skill Factory

Core Features

1. Information Retrieval: Get commit hash via

   git ls-remote

   , fetch README via HTTP
2. Directory Generation: Create standardized Skill directory structure (SKILL.md, scripts/, references/, assets/)
3. Metadata Injection: Automatically populate extended metadata fields (source_url, source_hash, version, etc.)
4. Placeholder Script: Generate wrapper.py template for subsequent implementation

Usage Scenarios

• /skill-factory <github_url>

• "Package this repository into a Skill: <url>"
• "Create a Skill from GitHub: <url>"
• "Wrap this tool: <url>"

Workflow

Step 1: Retrieve Repository Information

scripts/fetch_github_info.py

python scripts/fetch_github_info.py <github_url>

{
  "name": "repo-name",
  "owner": "owner",
  "url": "https://github.com/owner/repo",
  "latest_hash": "abc123...",
  "default_branch": "main",
  "description": "Repository description",
  "readme": "README content..."
}

Step 2: Analyze README

• Main functions of the tool
• Installation and usage methods
• CLI parameters or API interfaces
• Dependency requirements

Step 3: Create Skill Directory

scripts/create_skill.py

python scripts/create_skill.py <json_file_or_string> <output_dir>

skill-name/
├── SKILL.md              # Main Skill file (with extended metadata)
├── scripts/
│   └── wrapper.py        # Placeholder script
├── references/
│   └── .gitkeep
└── assets/
    └── .gitkeep

Step 4: Refine Skill

1. Refines the description and usage instructions in SKILL.md
2. Implements the actual logic in wrapper.py
3. Adds necessary reference documents

Step 5: Verification

• The frontmatter format of SKILL.md is correct
• source_hash is recorded correctly
• description includes trigger conditions
• wrapper.py is executable (if implemented)

Generated Metadata Format

---
name: skill-name
description: Detailed description including trigger conditions

# Lifecycle management fields (required)
source_url: https://github.com/owner/repo
source_hash: abc123def456...
version: 0.1.0
created_at: 2026-01-23
updated_at: 2026-01-23
evolution_enabled: true

# Optional fields
entry_point: scripts/wrapper.py
dependencies:
  - dependency1
  - dependency2
---

Script Explanation

scripts/fetch_github_info.py

scripts/create_skill.py

scripts/import_github_skill.py

import_github_skill.py (Recommended)

• Retrieve directory structure using GitHub API (single request)
• Parallel download of all files (10 threads)
• Automatically switch to git clone fallback when API is rate-limited
• Support custom Skill names
• Support removing source information (

  --no-source

  )

# Basic usage
python scripts/import_github_skill.py <github_url> <output_dir>

# Custom name
python scripts/import_github_skill.py <github_url> <output_dir> --name my-skill

# Do not retain source information (as a new local Skill)
python scripts/import_github_skill.py <github_url> <output_dir> --no-source

# Combination usage
python scripts/import_github_skill.py https://github.com/user/repo ./skills --name my-skill --no-source

# Import and rename
python scripts/import_github_skill.py https://github.com/PenglongHuang/chinese-novelist-skill ./skills --name chinese-novelist --no-source

Best Practices

1. Progressive Disclosure: Do not stuff the entire repository content into the Skill, only include necessary wrapper code
2. Dependency Isolation: Generated Skills should explicitly declare dependencies, recommend using venv or uv for management
3. Idempotency: The

   source_hash

   field allows skill-manager to check for updates later
4. Complete Description: The description must include "what it does" and "when to use"

Examples

Example 1: Wrap yt-dlp

User: /skill-factory https://github.com/yt-dlp/yt-dlp

Agent:
1. Run fetch_github_info.py to retrieve repository information
2. Analyze README to understand yt-dlp is a video download tool
3. Run create_skill.py to create directory
4. Refine SKILL.md and add common command examples
5. Implement wrapper.py to encapsulate download functionality

Example 2: Wrap ffmpeg-python

User: Package https://github.com/kkroening/ffmpeg-python into a Skill

Agent:
1. Retrieve repository information
2. Analyze README to understand this is Python bindings for FFmpeg
3. Create Skill directory
4. Refine documentation and add common conversion examples
5. Implement wrapper.py to encapsulate common operations

Collaboration with Other Skills

• skill-manager: Use source_url and source_hash to check for updates
• skill-evolution: Use evolution_enabled to control whether user experiences are recorded
• skill-creator: Follow the same directory structure and metadata specifications