nullclaw skills

Discover, install, and manage skill packs

The nullclaw skills command manages skill packs - reusable instruction sets and tool configurations that extend the agent’s capabilities.

Subcommands

nullclaw skills list              # List installed skills
nullclaw skills install <name>    # Install a skill pack
nullclaw skills remove <name>     # Remove a skill pack
nullclaw skills info <name>       # Show skill details

What are Skills?

Skills are packaged instruction sets that provide:

• Domain expertise: Specialized knowledge (medical, legal, coding, etc.)
• Tool configurations: Pre-configured tool chains
• Prompt templates: Reusable prompt patterns
• Context files: Reference documentation and examples

Skills are stored in ~/.nullclaw/workspace/skills/<name>/ with:

• SKILL.md - Main skill instructions
• skill.toml - Metadata and dependencies
• Supporting files (examples, references, scripts)

skills list

List all installed skills:

nullclaw skills list

Example Output

Installed Skills:

✓ python-expert
  Description: Expert Python development with best practices
  Version: 1.2.0
  Source: github.com/skillhub/python-expert
  Tools: shell, file_read, file_write

✓ devops-automation
  Description: DevOps automation and deployment
  Version: 2.0.1
  Source: github.com/skillhub/devops
  Tools: shell, git_operations, docker

✓ data-analysis
  Description: Data analysis with pandas and visualization
  Version: 1.0.0
  Source: local
  Tools: shell, file_read, browser

skills install

Install a skill pack:

# Install from GitHub
nullclaw skills install github.com/skillhub/python-expert

# Install from URL
nullclaw skills install https://example.com/skills/medical.zip

# Install from local path
nullclaw skills install ~/my-skills/custom-skill

Installation Process

1. Download or copy

   Skill pack is downloaded from source or copied from local path

2. Validate

   Check for required files:

   • SKILL.md (required)
   • skill.toml (optional metadata)

   Validate TOML schema if present

3. Install

   Copy to ~/.nullclaw/workspace/skills/<name>/

4. Register

   Add to skills registry for auto-loading

skills remove

Remove an installed skill:

nullclaw skills remove python-expert

This deletes the skill directory and updates the registry.

Removing a skill deletes all its files. Back up any custom modifications first.

skills info

Show detailed information about a skill:

nullclaw skills info python-expert

Example Output

Skill: python-expert
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Description: Expert Python development with best practices
Version: 1.2.0
Author: SkillHub Contributors
Source: github.com/skillhub/python-expert
License: MIT

Dependencies:
  - shell tool
  - file_read tool
  - file_write tool

Instructions (SKILL.md):
  - Type hints and docstrings required
  - Follow PEP 8 style guide
  - Use virtual environments
  - Write unit tests with pytest
  - Security: validate inputs, avoid eval()

Files:
  - SKILL.md (4.2 KB)
  - skill.toml (512 bytes)
  - examples/fastapi.py (1.8 KB)
  - examples/pytest.py (1.2 KB)

Skill Manifest Format

skill.toml

name = "python-expert"
version = "1.2.0"
description = "Expert Python development with best practices"
author = "SkillHub Contributors"
license = "MIT"
source = "github.com/skillhub/python-expert"

[dependencies]
tools = ["shell", "file_read", "file_write"]

[metadata]
tags = ["programming", "python", "backend"]
languages = ["python"]

SKILL.md

# Python Expert Skill

You are an expert Python developer. Follow these guidelines:

## Code Quality

- Always use type hints
- Write comprehensive docstrings
- Follow PEP 8 style guide
- Use meaningful variable names

## Best Practices

- Virtual environments for all projects
- requirements.txt or pyproject.toml for dependencies
- Unit tests with pytest
- Error handling with specific exceptions

## Security

- Validate all inputs
- Never use eval() or exec() with user input
- Use parameterized SQL queries
- Keep dependencies updated

## Examples

See `examples/` directory for reference implementations.

Skill Discovery

Skills can be discovered from:

• GitHub: Search for repositories tagged with nullclaw-skill
• ClawHub: Official skill registry (coming soon)
• HuggingFace: AI/ML focused skills
• Local: Create your own custom skills

Creating Custom Skills

1. Create directory

   mkdir -p ~/.nullclaw/workspace/skills/my-skill
   cd ~/.nullclaw/workspace/skills/my-skill
   
2. Write SKILL.md

   Create instructions for the agent:

   # My Custom Skill
   
   You are an expert in...
   
   ## Guidelines
   - ...
   
3. Optional: Add skill.toml

   name = "my-skill"
   version = "1.0.0"
   description = "My custom skill"
   
4. Test

   nullclaw skills info my-skill
   nullclaw agent -m "Use my-skill to..."
   

Skill Loading

Skills are automatically loaded when:

1. Agent starts
2. Skill is mentioned in prompt (“use python-expert skill”)
3. Skill is set as default in config

Loaded skills inject their instructions into the agent’s system prompt.

Examples

Install Python skill

nullclaw skills install github.com/skillhub/python-expert

Use in conversation

nullclaw agent -m "Using python-expert skill, write a FastAPI app with authentication"

List all skills

nullclaw skills list

Remove skill

nullclaw skills remove python-expert

Troubleshooting

Skill not loading

Error: Skill 'python-expert' not found

Fix: Verify installation:

nullclaw skills list
ls ~/.nullclaw/workspace/skills/

Invalid TOML

Error: Failed to parse skill.toml

Fix: Validate TOML syntax:

cat ~/.nullclaw/workspace/skills/my-skill/skill.toml
# Check for syntax errors

Missing dependencies

Warning: Skill requires 'composio' tool which is not available

Fix: Install required tools or remove dependency from skill.toml

See Also

• Agent - Use skills in conversations
• Tools - Available tools for skills
• Configuration - Configure default skills