Outclaw

CLI tool for managing Claude Code Skills - A complete registry system.

Installation

npm install -g outclaw

Usage

Create a new skill

# Interactive mode
outclaw init

# With name
outclaw init my-skill

# Create in global directory
outclaw init my-skill --global

Install a skill

# From GitHub
outclaw install github:owner/repo
outclaw install github:owner/repo@branch

# From GitHub (shorthand)
outclaw install owner/repo

# From URL
outclaw install https://github.com/owner/repo

# Install globally
outclaw install github:owner/repo --global

List installed skills

# List all skills
outclaw list

# List only global skills
outclaw list --global

# List only project skills
outclaw list --project

# Output as JSON
outclaw list --json

Search for skills

outclaw search <query>

# Limit results
outclaw search react --limit 10

Show skill info

outclaw info my-skill

Uninstall a skill

outclaw uninstall my-skill

# Uninstall global skill
outclaw uninstall my-skill --global

Claude Code Skills Format

Skills follow the Agent Skills open standard.

Directory Structure

my-skill/
├── SKILL.md           # Main instructions (required)
├── references/        # Reference documentation (optional)
├── examples/          # Example outputs (optional)
└── scripts/           # Executable scripts (optional)

SKILL.md Format

---
name: my-skill
description: What this skill does and when to use it
version: 1.0.0
disable-model-invocation: false
user-invocable: true
allowed-tools: Read, Grep, Glob
argument-hint: [file-path]
---

Your skill instructions here...

Frontmatter Options

| Field | Description | |-------|-------------| | name | Skill name (lowercase, alphanumeric, hyphens) | | description | What the skill does (used for trigger matching) | | version | Semantic version | | disable-model-invocation | Prevent automatic triggering | | user-invocable | Allow manual invocation via /skill-name | | allowed-tools | Tools that can be used without permission | | argument-hint | Hint for autocomplete | | model | Override model for this skill | | context | Run in fork for isolated subagent | | agent | Subagent type when using context: fork |

Skills Location

• Global: ~/.claude/skills/<skill-name>/SKILL.md
• Project: .claude/skills/<skill-name>/SKILL.md

Using as a Claude Code Skill

You can create a skill to help manage skills:

---
name: outclaw
description: Manage Claude Code skills - create, install, search, and publish
disable-model-invocation: true
allowed-tools: Bash
---

Use outclaw CLI to manage Claude Code skills:

- Create new skill: `outclaw init <name>`
- Install skill: `outclaw install <skill>`
- List skills: `outclaw list`
- Search skills: `outclaw search <query>`

License

MIT