FastSkill

Package manager and operational toolkit for Agent AI Skills. FastSkill enables discovery, installation, versioning, and deployment of skills at scale.

Add a skill

fastskill add https://github.com/org/skill-repo

Quickly add skills from Git, local folders, or registries to extend your AI agent's capabilities.

What is FastSkill?

FastSkill is a skill package manager and operational toolkit for the AI agent ecosystem. It builds on Anthropic's standardized Skills format so you can install, validate, version, and ship skills the same way you manage other dependencies: manifests and lockfiles, fastskill eval for quality checks, semantic and catalog search, optional registry publish, plus a local HTTP API and UI when you run fastskill serve.

What are skills? Skills are reusable instruction sets in SKILL.md that extend an AI agent's capabilities with specialized procedures, tool integrations, and domain knowledge. FastSkill focuses on managing that corpus: what is installed, whether metadata is valid, how versions move, and how teams keep agents in sync—not on executing skill code inside FastSkill itself.

AI Agentic Skills Standard

FastSkill supports the AI Agentic Skills Specification.

Skill Requirements

According to the standard, a skill must have:

Required: SKILL.md file with YAML frontmatter
- name: Skill identifier (1-64 chars, lowercase alphanumeric + hyphens)
- description: What the skill does (1-1024 chars)
Optional: skill-project.toml for advanced features
- Recommended for skill authors
- Provides dependency management
- Enables version tracking

Adding Skills Without skill-project.toml

# Works with standard-compliant skills
fastskill add ./my-skill

# Will use metadata from SKILL.md:
# - Skill ID: from metadata.id or name field
# - Version: from metadata.version or defaults to 1.0.0
# - Warning displayed recommending 'fastskill init'

Migration Guide

For existing skills without skill-project.toml:

# Navigate to skill directory
cd my-skill

# Run fastskill init to add skill-project.toml
fastskill init

# This creates skill-project.toml with:
# [metadata]
# id = "my-skill"  # From SKILL.md name
# version = "1.0.0"  # From SKILL.md metadata.version

Key Capabilities

Package management: Install, update, and remove skills from multiple sources (Git, local, ZIP, registries)
Manifests & locks: Declarative skill-project.toml and skills.lock for reproducible installs and groups
Validation & reconciliation: Catch bad metadata and drift between manifest, lock, and installed skills
Evaluations: fastskill eval to validate and run structured checks against skills (see fastskill eval --help)
Semantic search: Find installed skills with embeddings and natural language; catalog search for remote discovery
Agent sync: fastskill sync writes installed skills into agent metadata files (Claude Code–compatible layouts)
Publish & registries: Package ZIPs and publish to an API or storage when you distribute skills to others
HTTP API & Web UI: Optional local fastskill serve for browsing and integration (unauthenticated by default)

Compatible Agents

Optimized for Claude Code. Skills follow the same SKILL.md format used by Cursor and other AI agents.

Core Use Cases

Skill authors: fastskill init, package, validate, and run evals before you publish
Agent developers: Install, list, show, read, and search skills your agents actually load
Teams: One manifest and lockfile, optional groups, private sources when you share internally
CI/CD: Reproducible install --lock, packaging, and publish steps in pipelines
Production systems: Operate skill sets at scale—updates, rollbacks, and drift checks—not ad-hoc copies

Benefits Over Baseline

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: Handles large skill trees, many projects, and multiple sources without manual bookkeeping
Automation: CLI and API enable CI/CD integration

Installation

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

CLI Installation

Quick Install (Recommended)

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 v0.6.8)
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 (macOS + Linux)

Install FastSkill via Homebrew on macOS or 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.

Manual Installation from GitHub Releases

Download the pre-built binary for your platform from GitHub Releases.

macOS:

Two macOS binaries are available:

Binary	Hardware
`fastskill-aarch64-apple-darwin.tar.gz`	Apple Silicon (M1/M2/M3+)
`fastskill-x86_64-apple-darwin.tar.gz`	Intel Macs

Apple Silicon example:

VERSION="0.8.6"  # Replace with latest version
wget https://github.com/gofastskill/fastskill/releases/download/v${VERSION}/fastskill-aarch64-apple-darwin.tar.gz
tar -xzf fastskill-aarch64-apple-darwin.tar.gz
sudo mv fastskill /usr/local/bin/
fastskill --version

Intel macOS example:

VERSION="0.8.6"  # Replace with latest version
wget https://github.com/gofastskill/fastskill/releases/download/v${VERSION}/fastskill-x86_64-apple-darwin.tar.gz
tar -xzf fastskill-x86_64-apple-darwin.tar.gz
sudo mv fastskill /usr/local/bin/
fastskill --version

Linux:

Two Linux binaries are available:

Binary	Best For	Compatibility
`fastskill-x86_64-unknown-linux-musl.tar.gz`	Containers, CI/CD, older distributions	Universal - works on any Linux (Ubuntu 18.04+, RHEL 7+, Alpine, etc.). Note: Built without git-support; use system git for git operations.
`fastskill-x86_64-unknown-linux-gnu.tar.gz`	FIPS/compliance, full git2 support	Requires glibc 2.38+ (Ubuntu 24.04+, Fedora 39+). Includes native git integration.

Recommended: Use the musl (static) binary for maximum compatibility:

VERSION="0.8.6"  # Replace with latest version
wget https://github.com/gofastskill/fastskill/releases/download/v${VERSION}/fastskill-x86_64-unknown-linux-musl.tar.gz
tar -xzf fastskill-x86_64-unknown-linux-musl.tar.gz
sudo mv fastskill /usr/local/bin/
fastskill --version

For FIPS/compliance environments requiring dynamic linking:

VERSION="0.8.6"  # Replace with latest version
wget https://github.com/gofastskill/fastskill/releases/download/v${VERSION}/fastskill-x86_64-unknown-linux-gnu.tar.gz
tar -xzf fastskill-x86_64-unknown-linux-gnu.tar.gz
sudo mv fastskill /usr/local/bin/
fastskill --version

From Source:

cargo install fastskill
# Or build from source
git clone https://github.com/gofastskill/fastskill.git
cd fastskill
cargo install --path .

Requirements: Rust nightly (for source builds), OpenAI API key for embedding features

Kubernetes Deployment (Production)

Deploy FastSkill as a production service in Kubernetes using Helm:

# Create secrets and install chart
kubectl create namespace fastskill
kubectl create secret generic fastskill-github-token \
  --from-literal=GITHUB_TOKEN=your-token -n fastskill
kubectl create secret generic fastskill-s3-credentials \
  --from-literal=AWS_ACCESS_KEY_ID=your-key \
  --from-literal=AWS_SECRET_ACCESS_KEY=your-secret -n fastskill

helm install fastskill ./tools/fastskill/helm/fastskill \
  --namespace fastskill --create-namespace

For detailed Kubernetes deployment instructions, see the Kubernetes Deployment Guide.

Quick Start

1. Configure FastSkill

Create skill-project.toml in your project root:

[metadata]
id = "my-project"
version = "1.0.0"

[dependencies]
# Add skill dependencies here

[tool.fastskill]
skills_directory = ".claude/skills"

[tool.fastskill.embedding]
openai_base_url = "https://api.openai.com/v1"
embedding_model = "text-embedding-3-small"

[[tool.fastskill.repositories]]
name = "anthropic"
type = "git-marketplace"
url = "https://github.com/anthropics/skills"
priority = 0

Or use the init command:

fastskill init

Set your OpenAI API key:

export OPENAI_API_KEY="your-key-here"

2. Add Skills

Source formats

Source	Example
Git URL	`https://github.com/org/skill.git`
Tree URL (subdir)	`https://github.com/org/repo/tree/main/path/to/skill`
Local path	`./local-skill`
Recursive directory	`./skills -r`
Editable (dev)	`./local-skill -e`

Options

Flag	Description
`-r, --recursive`	Add all skills under directory (local folders only)
`-e, --editable`	Install in editable mode for local development
`-f, --force`	Force registration even if skill exists
`--branch <BRANCH>`	Git branch to checkout (git URLs only)
`--tag <TAG>`	Git tag to checkout (git URLs only)
`--source-type <TYPE>`	Override source type (registry, github, local)
`--group <GROUP>`	Add skill to a specific group

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

# Add skill from a subdirectory (GitHub tree URL: tree/branch/path/to/skill)
fastskill add "https://github.com/org/repo/tree/main/path/to/skill"

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

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

# Add all skills under a directory (recursive)
fastskill add ./skills -r

3. Index and Search

# Index skills for semantic search
fastskill reindex

# Search for skills (remote catalog by default)
fastskill search "powerpoint presentation"
fastskill search "data processing" --limit 5

# Search local/installed skills only
fastskill search "pptx" --local

# Search using keyword-only (no API key required)
fastskill search --embedding false "powerpoint"

# Search specific repository
fastskill search "text processing" --repo my-repo

Essential Commands

Command	Description
`fastskill add <source>`	Add skill from Git, local, or registry
`fastskill remove <skill-id>`	Remove skill from database
`fastskill show`	List installed skills and metadata
`fastskill update`	Update skills to latest versions
`fastskill search "query"`	Search skills (remote catalog by default, use --local for installed)
`fastskill reindex`	Rebuild vector index for search
`fastskill serve`	Start HTTP API server
`fastskill init`	Initialize skill-project.toml
`fastskill package`	Package skills into ZIP artifacts
`fastskill analyze matrix`	Show similarity matrix between all skills

Skill Management

fastskill add <source>              # Add skill (git URL, tree URL for subdir, local path, or ZIP). Use -r for recursive folder add
fastskill remove <skill-id>          # Remove skill
fastskill show                       # List installed skills
fastskill update                     # Update skills to latest versions

Search and Discovery

fastskill search "query"             # Search remote catalog (default)
fastskill search "query" --local     # Search installed/local skills
fastskill search "query" --repo my-repo  # Search specific repository
fastskill reindex                    # Rebuild search index

Server and API

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

Skill Authoring

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

Diagnostic Commands

fastskill analyze matrix             # Show similarity matrix between all skills

Identify related skills
Find potential duplicates
Verify embedding quality
Export data in JSON format

Repositories

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

Public registries (HTTP-based index with 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 skill-project.toml under [[tool.fastskill.repositories]].

Repository Management

Use the repos command to manage repositories and browse remote catalogs:

# List repositories
fastskill repos list

# Add a repository
fastskill repos add my-repo --repo-type local /path/to/skills

# Browse catalog skills
fastskill repos skills

# Show skill details
fastskill repos show pptx

# Show available versions
fastskill repos versions pptx

**Command Architecture**: - Use `fastskill repos` for repository management (add, remove, list, test, refresh) and catalog browsing (skills, show, versions) - Use `fastskill search` for searching skills (remote catalog by default, use --local for installed skills, --repo for specific repository)

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

Configuration

FastSkill uses skill-project.toml as the unified configuration file for both project-level and skill-level contexts.

skill-project.toml Structure

[metadata]
id = "my-skill"
version = "1.0.0"

[dependencies]
# Add your skill dependencies here

[tool.fastskill]
skills_directory = ".claude/skills"

[tool.fastskill.embedding]
openai_base_url = "https://api.openai.com/v1"
embedding_model = "text-embedding-3-small"

[[tool.fastskill.repositories]]
name = "anthropic"
type = "git-marketplace"
url = "https://github.com/anthropics/skills"
priority = 0

Configuration Resolution

The CLI resolves configuration from skill-project.toml:

Searches current directory and parents for skill-project.toml
Extracts [tool.fastskill] section for skills directory and embedding config
Extracts [tool.fastskill.repositories] for repository sources
Defaults to .claude/skills/ if no skills_directory is configured

Installation Scope

FastSkill is project-scoped. Configuration and skills are managed per project using skill-project.toml in the project root. There is no global/user-level mode; each project maintains its own skills directory and dependencies.

Skill Discovery

FastSkill looks for skills in the following locations:

skills_directory - Configured path in [tool.fastskill.skills_directory]
Default - .claude/skills/ if not specified
Repositories - All sources from [[tool.fastskill.repositories]] including:
- Git marketplace repos with marketplace.json
- HTTP registries with flat index
- ZIP URL sources with marketplace.json
- Local filesystem paths

Environment Variables

Variable	Description
`OPENAI_API_KEY`	OpenAI API key for semantic search and embeddings (required)
`RUST_LOG`	Logging level (e.g., `fastskill=debug`, `fastskill=trace`)
`FASTSKILL_API_URL`	Base URL for registry API
`FASTSKILL_API_TOKEN`	Authentication token for registry API
`FASTSKILL_CONFIG_DIR`	Path to FastSkill configuration directory
`FASTSKILL_STATIC_DIR`	Path to static files for HTTP server

Quick Setup

# Initialize a new project
fastskill init

# Set OpenAI API key for semantic search
export OPENAI_API_KEY="your-key-here"

Troubleshooting

Configuration Not Found

If you see "Embedding configuration required but not found":

Run fastskill init to create skill-project.toml
Add [tool.fastskill.embedding] section with embedding configuration
Set OPENAI_API_KEY environment variable

API Key Issues

export OPENAI_API_KEY="your-openai-api-key-here"

For persistent setup, add this to your shell profile (.bashrc, .zshrc, etc.).

Add fails (git or network)

If fastskill add fails with git or network errors:

Verify the Git URL is accessible and public (or configured with auth)
Check network connectivity and proxy settings
For private repos, ensure credentials are configured
Use fastskill add --verbose for detailed error messages

Search returns no results

If fastskill search --local returns no results:

Run fastskill reindex to rebuild the search index
Verify OPENAI_API_KEY is set and valid
Check embedding configuration in [tool.fastskill.embedding]
Ensure skills are installed (fastskill show)

If fastskill search (remote) returns no results:

Verify repositories are configured (fastskill repos list)
Check repository connectivity (fastskill repos test <repo-name>)
Refresh repository cache (fastskill repos refresh)

Documentation

Registry Setup - Detailed registry configuration and management
Kubernetes Deployment - Production deployment guide
Security Policy - Security guidelines and vulnerability reporting
GitHub Releases - Latest versions and changelog

License

Apache-2.0

Name		Name	Last commit message	Last commit date
Latest commit History 260 Commits
.github		.github
docs		docs
scripts		scripts
src		src
tests		tests
webdocs		webdocs
.gitignore		.gitignore
.repomixignore		.repomixignore
AGENTS.md		AGENTS.md
CONTRIBUTING.md		CONTRIBUTING.md
Cargo.lock		Cargo.lock
Cargo.toml		Cargo.toml
README.md		README.md
SECURITY.md		SECURITY.md
STYLE.md		STYLE.md
coverage-before.lcov		coverage-before.lcov
nextest.toml		nextest.toml
repomix.config.json		repomix.config.json
rust-toolchain.toml		rust-toolchain.toml

Folders and files

Latest commit

History

Repository files navigation

FastSkill

Add a skill

What is FastSkill?

AI Agentic Skills Standard

Skill Requirements

Adding Skills Without skill-project.toml

Migration Guide

Key Capabilities

Compatible Agents

Core Use Cases

Benefits Over Baseline

Installation

CLI Installation

Kubernetes Deployment (Production)

Quick Start

1. Configure FastSkill

2. Add Skills

3. Index and Search

Essential Commands

Skill Management

Search and Discovery

Server and API

Skill Authoring

Diagnostic Commands

Repositories

Repository Management

Configuration

skill-project.toml Structure

Configuration Resolution

Installation Scope

Skill Discovery

Environment Variables

Quick Setup

Troubleshooting

Configuration Not Found

API Key Issues

Add fails (git or network)

Search returns no results

Documentation

License

CI Test

About

Resources

Contributing

Security policy

Uh oh!

Stars

Watchers

Forks

Releases 98

Uh oh!

Contributors

Uh oh!

Languages