Package manager and operational toolkit for Claude Code-compatible skills. FastSkill enables discovery, installation, versioning, and deployment of skills at scale.

FastSkill is a skill package manager and operational toolkit for the AI agent ecosystem. It builds on Anthropic's standardized Skills format, adding registry services, semantic search, version management, and deployment tooling.

Skills are recipes that extend AI Agent capabilities with specialized procedures, tool integrations, and domain knowledge. FastSkill provides the infrastructure to develop, manage, consume, and deploy skills at scale.

• Package Management: Install, update, and remove skills from multiple sources (Git, local, ZIP)
• Semantic Search: Find skills using OpenAI embeddings and natural language queries
• Registry Services: Publish, version, and distribute skills via registry
• Manifest System: Declarative dependency management with lock files for reproducible installations
• HTTP API: RESTful service layer for agent integration
• Web UI: Browse and manage skills through web interface

• Skill Authors: Package, version, and publish skills to registries
• Agent Developers: Discover and install skills for agent capabilities
• Teams: Share internal skills via private registries with version control
• CI/CD: Automate skill packaging, publishing, and deployment pipelines
• Production Systems: Manage skill lifecycles in agentic applications

Without FastSkill, teams manually manage skills through ad-hoc scripts, copy-paste workflows, and manual version tracking. FastSkill provides:

• Standardized Workflows: Consistent patterns for skill lifecycle management
• Version Control: Semantic versioning and dependency resolution
• Discovery: Semantic search eliminates manual skill cataloging
• Reproducibility: Lock files ensure consistent installations across environments
• Scalability: Registry architecture supports hundreds of skills across teams
• Automation: CLI and API enable CI/CD integration

FastSkill can be installed in several ways depending on your use case:

Quick Install

Install FastSkill with a single command:

curl -fsSL https://raw.githubusercontent.com/gofastskill/fastskill/main/scripts/install.sh | bash

Or download and run the script manually:

wget https://raw.githubusercontent.com/gofastskill/fastskill/main/scripts/install.sh
chmod +x install.sh
./install.sh

The script automatically:

• Detects your platform
• Downloads the latest version (or specify a version: ./install.sh v0.7.8)
• Installs to /usr/local/bin (or ~/.local/bin if sudo is unavailable)
• Verifies the installation

Options:

• --user: Install to ~/.local/bin instead of system directory
• --prefix DIR: Install to a custom directory
• --force: Overwrite existing installation
• --help: Show all available options

Homebrew (Linux)

Install FastSkill via Homebrew on Linux:

brew install gofastskill/cli/fastskill

For more details, see the Homebrew tap repository.

Scoop (Windows)

Install FastSkill via Scoop on Windows:

scoop bucket add gofastskill https://github.com/gofastskill/scoop-bucket
scoop install fastskill

For more details, see the Scoop bucket repository.

Create .fastskill.yaml in your project root:

embedding:
  openai_base_url: "https://api.openai.com/v1"
  embedding_model: "text-embedding-3-small"

Set your OpenAI API key:

export OPENAI_API_KEY="your-key-here"

# Add skill from git URL
fastskill add https://github.com/org/skill.git

# Add skill from local folder
fastskill add ./local-skill

# Add skill in editable mode (for local development)
fastskill add ./local-skill -e

# Index skills for semantic search
fastskill reindex

# Search for skills
fastskill search "powerpoint presentation"
fastskill search "data processing" --limit 5

fastskill add <source>              # Add skill (git URL, local path, or ZIP)
fastskill remove <skill-id>          # Remove skill
fastskill show                       # List installed skills
fastskill update                     # Update skills to latest versions

fastskill search "query"             # Semantic search
fastskill reindex                    # Rebuild search index

fastskill serve                      # Start HTTP API server
fastskill serve --enable-registry    # Enable web UI at /registry
fastskill proxy                      # Start Claude API proxy server

fastskill init                       # Initialize skill metadata
fastskill package                    # Package skills into ZIP artifacts
fastskill package --force --recursive # Package all skills recursively from nested directories

FastSkill provides a unified repository system for managing all skill storage locations. Repositories can be:

• Public registries (crates.io-style with Git index + S3 storage)
• Private registries (enterprise/internal registries)
• Git repositories (with marketplace.json for skill discovery)
• ZIP URL sources (static hosting with marketplace.json)
• Local folders (for development)

All repository types are configured in .claude/repositories.toml (or automatically loaded from sources.toml and registries.toml for backward compatibility).

For detailed repository setup, usage, and management instructions, see docs/REGISTRY.md.

Create .fastskill.yaml in your project root:

embedding:
  openai_base_url: "https://api.openai.com/v1"
  embedding_model: "text-embedding-3-small"

# Optional: Custom skills directory
skills_directory: ".claude/skills"

Apache-2.0