neoskills

Homebrew-style skill manager for AI coding agents.

neoskills manages portable skill definitions across agent ecosystems (Claude Code, OpenCode, OpenClaw). Browse skills in one place, sync to GitHub, deploy selectively via symlinks, and discover relationships through a built-in ontology graph.

---

Table of Contents

• Installation
• Quick Start
• How It Works
• Ontology Layer
• Agent Targets
• Operating Modes
• CLI Reference
• Development

---

Installation

From PyPI (recommended)

pip install neoskills

Or with uv:

uv tool install neoskills

From source

git clone https://github.com/neolaf2/neoskills
cd neoskills
uv sync --dev
uv run neoskills --help

Requirements

• Python 3.13+ (required)
• Git (for tap operations)
• uv (recommended for development)

---

Quick Start

# 1. Initialize the workspace
neoskills init

# 2. Add a tap (a git-hosted skill repository)
neoskills tap https://github.com/your-org/my-skills

# 3. Browse and search
neoskills list                          # all skills in your taps
neoskills list --linked                 # what's deployed to your agent
neoskills search "document processing"  # cross-tap search

# 4. Deploy a skill to your agent
neoskills install kstar-loop            # copies to tap + symlinks to agent

# 5. Create your own skill
neoskills create my-new-skill -d "What it does" --type task

# 6. Check system health
neoskills doctor

What just happened?

neoskills init created ~/.neoskills/ with a default tap. install found the skill in a tap and created a symlink into your agent’s skill directory (e.g., ~/.claude/skills/). Your agent can now use the skill. create scaffolded a new skill with SKILL.md + ontology.yaml.

---

How It Works

Architecture

~/.neoskills/                         # Workspace root
├── config.yaml                       # Targets, taps, defaults
├── taps/                             # Git-cloned skill repositories
│   ├── mySkills/                     # Default tap
│   │   └── skills/
│   │       └── /
│   │           ├── SKILL.md          # Skill definition (frontmatter + body)
│   │           ├── ontology.yaml     # Graph metadata (optional, recommended)
│   │           ├── scripts/          # Executable code (optional)