Skip to main content

AI Agent Skill Search and Management

Project description

skill

AI Agent Skill Search and Management

Manage Agent Skills (SKILL.md files) across multiple AI coding agents — Claude Code, Cursor, GitHub Copilot, Windsurf, and more.

Install

pip install skill

Quick Start

from skill import search, create, install, validate

# Create a skill locally
skill = create('my-skill', description='My custom coding rules')

# Validate it
issues = validate('/path/to/skill-dir')

# Search for skills
results = search('react best practices', local_only=True)

# Install to an agent target
install('_local/my-skill', agent_targets=['claude-code'], scope='project')

Link skills from a project

If a project ships skills (in .claude/skills/, {pkg}/data/skills/, or a dedicated folder), you can symlink them all into your agent's skills directory in one shot:

from skill import link_skills

# Point at a project root — it finds the skills automatically
link_skills('/path/to/my-project')

# Or point at the skills folder directly
link_skills('/path/to/my-project/my_pkg/data/skills')

# Symlink into a specific target instead of ~/.claude/skills
link_skills('/path/to/my-project', target='/other/project/.claude/skills')

Each skill is validated before linking — invalid skills are skipped with a warning. The target directory is also checked to ensure it's a recognized skills directory.

CLI

skill create my-skill --description "My custom coding rules"
skill search "react best practices" --local-only
skill validate ./my-skill/
skill list-skills
skill install _local/my-skill --agent-targets claude-code
skill link-skills /path/to/project
skill link-skills /path/to/project --target ~/.claude/skills --force

Shell completion

Enable tab completion for all skill commands:

skill install-completion

This detects your shell (bash/zsh) and adds the registration line to your shell config (~/.bashrc or ~/.zshrc). Restart your shell or source the config file to activate.

If you prefer to set it up manually:

# Add to your shell config:
eval "$(register-python-argcomplete skill)"

The first time you run any skill command without completion set up, you'll see a one-time hint reminding you to run skill install-completion.

Skills

This package ships its own AI agent skills (in skills/), installable into any agent host with gh skill (GitHub CLI ≥ 2.90.0):

gh skill preview thorwhalen/skill skill-package-setup        # inspect before installing
gh skill install thorwhalen/skill skill-package-setup --agent claude-code
gh skill install thorwhalen/skill skill-enable
Skill What it covers
skill-package-setup Canonical layout & distribution policy — where a package's skills live, gh skill setup, and migration
skill-build Author skills for a package by analyzing its code, tests, and docs
skill-enable Ship skills via pip ({pkg}/data/skills/) + audience classification
skill-docs Document a project's skills in its README
skill-manage Search, list, inspect, install/uninstall skills across agents
skill-sync Keep skills in sync with the code they document

These are dev/maintainer skills, so they're not bundled in the wheel — install them with gh skill as above. See skill-package-setup for the layout policy this repo follows (real files in skills/, relative symlinks in .claude/skills/).

Plugin System

skill has a registry-based plugin architecture with four extension points:

Registry What it holds Import
Agent targets Where agents expect skills installed from skill.install import agent_targets
Translators SKILL.md → target format converters from skill.translate import translators
Backends Remote skill sources for search from skill.search import backends
Validators Pluggable validation rules from skill.create import validators

Register at runtime:

from skill.install import agent_targets, AgentTarget

agent_targets.register('windsurf', AgentTarget(
    name='windsurf',
    project_path='{project}/.windsurf/rules/{name}.md',
    format='skill.md',
))

Or via entry points in your pyproject.toml:

[project.entry-points."skill.agent_targets"]
windsurf = "my_package:windsurf_target"

See Plugin System Documentation for full details on contracts, interfaces, and built-in registrations.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

skill-0.1.10.tar.gz (155.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

skill-0.1.10-py3-none-any.whl (42.4 kB view details)

Uploaded Python 3

File details

Details for the file skill-0.1.10.tar.gz.

File metadata

  • Download URL: skill-0.1.10.tar.gz
  • Upload date:
  • Size: 155.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.11.23 {"installer":{"name":"uv","version":"0.11.23","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

File hashes

Hashes for skill-0.1.10.tar.gz
Algorithm Hash digest
SHA256 70d74afdf89434cfccdb9110e7d559e340d39e760e6eca59465e4f3b5c59e485
MD5 dc6befc8ea9d4eb8e130df01a4c9d36e
BLAKE2b-256 73a38c279312e96a3d295d07938b97a29dbd4a66d026bd3d168f0ea8cb1017c1

See more details on using hashes here.

File details

Details for the file skill-0.1.10-py3-none-any.whl.

File metadata

  • Download URL: skill-0.1.10-py3-none-any.whl
  • Upload date:
  • Size: 42.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.11.23 {"installer":{"name":"uv","version":"0.11.23","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

File hashes

Hashes for skill-0.1.10-py3-none-any.whl
Algorithm Hash digest
SHA256 817feff028fbc30591a4d3982fe291916940d92a7ffa309201a536451aa4124c
MD5 796e385ac4969df681061868e19f71f2
BLAKE2b-256 078ea3f78c02524f2f6d1b8b163c96cda308ea5a6f09e29e7175ce3a29883a65

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page