agentskill-cli

AgentSkill 🛠️

Universal AI Skill Manager — Create, manage, and share coding rules for AI assistants like GitHub Copilot, Claude Code, and Cursor.

---

🎯 What is AgentSkill?

AgentSkill is a CLI tool that helps you create and manage skills (coding rules and best practices) that AI coding assistants follow when helping you write code.

The Problem

• You want AI to follow your coding standards consistently
• You have to repeat the same instructions in every project
• Different AI tools (Copilot, Claude, Cursor) need different file formats
• Sharing coding standards with your team is difficult

The Solution

AgentSkill lets you:

• ✅ Create skills once, use everywhere
• ✅ Install skills in any project with one command
• ✅ Auto-detect and support multiple AI tools
• ✅ Share skills with your team via export/import
• ✅ Update skills across all projects instantly

---

🚀 Installation

pip install agentskill-cli

After installation, both agentskill and ask (short) commands are available.

---

📖 Use Cases

1. Enforce Coding Standards

Create a skill to enforce your team's coding standards:

agentskill create python-standards -l python

Edit the skill to add your rules:

## Python Standards
- Use type hints for all function parameters and return values
- Use `pathlib` instead of `os.path`
- Prefer f-strings over .format()
- All functions must have docstrings
- Maximum line length: 88 characters (Black formatter)

2. Framework-Specific Rules

Create skills for specific frameworks with built-in templates:

# React component rules
agentskill create react-rules -l javascript -f react

# Django API rules  
agentskill create django-api -l python -f django

# Flutter widget rules
agentskill create flutter-app -l dart -f flutter

3. Project-Specific Instructions

Create skills for specific project requirements:

agentskill create myproject-rules

Add project-specific rules:

## MyProject Rules
- Use PostgreSQL for all database operations
- All API endpoints must be versioned (/api/v1/...)
- Use Redis for caching
- JWT tokens for authentication
- Follow company naming conventions

4. Team Collaboration

Share skills across your team:

# Export skill
agentskill export react-rules
# Creates: react-rules.skill.zip

# Team member imports
agentskill import react-rules.skill.zip

5. Multi-Project Consistency

Install the same skill in multiple projects:

cd ~/project-1
agentskill install python-standards

cd ~/project-2  
agentskill install python-standards

cd ~/project-3
agentskill install python-standards

Update skill once, sync everywhere:

agentskill edit python-standards  # Make changes
agentskill update python-standards  # Sync to all projects

---

📋 Quick Start

Step 1: Create a Skill

agentskill create my-rules -l python -f django

Step 2: Edit the Skill (optional)

agentskill edit my-rules

Step 3: Install in Your Project

cd /path/to/your/project
agentskill install my-rules

Step 4: Done!

AI assistants will now follow your rules when working in this project.

---

🔧 Commands

---

🎨 Creating Skills

Basic Skill

agentskill create my-rules

With Language & Framework Templates

# Python + Django
agentskill create django-api -l python -f django

# JavaScript + React
agentskill create react-component -l javascript -f react

# TypeScript + Next.js
agentskill create nextjs-app -l javascript -f nextjs

# Dart + Flutter
agentskill create flutter-widget -l dart -f flutter

# Go + Gin
agentskill create go-api -l go -f gin

# Rust + Actix
agentskill create rust-api -l rust -f actix

Supported Templates

---

📁 Skill Structure

my-skill/
├── SKILL.md          # Main instructions (loaded by AI)
├── skill.yaml        # Metadata (language, framework, version)
├── instructions.md   # Legacy format for compatibility
├── scripts/          # Helper scripts
├── references/       # Additional documentation
│   └── examples.md   # Code examples
└── assets/           # Templates, boilerplate code

SKILL.md Format

---
name: my-skill
description: 'Rules for Python Django development'
---

# My Skill

## When to Use
- Working on Django projects
- Building REST APIs

## Rules
- Use class-based views for complex logic
- Always use Django ORM
- Keep business logic in services

## Examples
Check `./references/examples.md` for code examples.

---

🔌 Supported AI Tools

Install for Specific Tool

# GitHub Copilot only
agentskill install my-skill -t copilot

# Claude only
agentskill install my-skill -t claude

# Cursor only
agentskill install my-skill -t cursor

# All tools
agentskill install my-skill -t all

---

🏗️ How It Works

┌─────────────────────────────────────────────────────────────────┐
│                        AgentSkill Flow                          │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  1. CREATE                    2. STORE                          │
│  ┌─────────────────┐         ┌─────────────────┐               │
│  │ agentskill      │         │ ~/.agentskill/  │               │
│  │ create my-skill │ ──────▶ │   skills/       │               │
│  └─────────────────┘         │     my-skill/   │               │
│                              │       SKILL.md  │               │
│                              └─────────────────┘               │
│                                      │                          │
│  3. INSTALL                          │                          │
│  ┌─────────────────┐                 │                          │
│  │ agentskill      │                 ▼                          │
│  │ install my-skill│         ┌─────────────────┐               │
│  └─────────────────┘         │ your-project/   │               │
│                              │   .github/      │               │
│         COPY ───────────────▶│     skills/     │               │
│                              │       my-skill/ │               │
│                              └─────────────────┘               │
│                                      │                          │
│  4. AI READS                         ▼                          │
│  ┌─────────────────┐         ┌─────────────────┐               │
│  │ GitHub Copilot  │◀────────│ Follows your    │               │
│  │ Claude Code     │         │ coding rules!   │               │
│  │ Cursor          │         └─────────────────┘               │
│  └─────────────────┘                                            │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

---

💡 Tips & Best Practices

Writing Effective Skills

1. Be Specific: Use clear, actionable rules
2. Add Examples: Include code examples in references/examples.md
3. Keep It Focused: One skill per domain/framework
4. Use Trigger Words: AI finds skills by matching keywords in descriptions

Skill Organization

# Recommended skill structure
agentskill create company-standards     # General company rules
agentskill create python-backend        # Python backend rules
agentskill create react-frontend        # React frontend rules
agentskill create api-design            # API design rules

Team Workflow

# 1. Create shared skills repository
mkdir team-skills && cd team-skills
git init

# 2. Export skills
agentskill export python-standards
agentskill export react-rules

# 3. Commit and push
git add .
git commit -m "Add team skills"
git push

# 4. Team members import
git clone <repo>
agentskill import python-standards.skill.zip
agentskill import react-rules.skill.zip

---

🔄 Updating Skills

Edit and Update

# Edit the skill
agentskill edit my-skill

# Update in all projects where it's installed
agentskill update my-skill

Version Tracking

Skills have version numbers in skill.yaml:

name: my-skill
version: 1.0

---

🗑️ Managing Skills

Delete a Skill

agentskill delete my-skill

# Force delete (no confirmation)
agentskill delete my-skill -f

View Storage Location

agentskill where
# Output: ~/.agentskill/skills/

---

📦 Sharing Skills

Export

agentskill export my-skill
# Creates: my-skill.skill.zip

Import

agentskill import my-skill.skill.zip

---

📄 License

This project is licensed under the MIT License.

---

🙏 Acknowledgments

• Inspired by the need for consistent AI-assisted coding
• Built with Typer for the CLI
• Supports GitHub Copilot, Claude Code, and Cursor

---

📧 Author

• Parth Kher

---

Made with ❤️ for developers who want AI to follow their rules.