Documentation

Learn & Build

Everything you need to get started with AI agent plugins

Core Concepts

Understanding the building blocks of plugin development

🤖

Agents

Specialized AI assistants with specific expertise. Each agent has a defined role, personality, and set of capabilities.

Example: Python Agent

agents/python-expert.md
├── Name: python-expert
├── Model: sonnet
├── Tools: Read, Write, Bash
└── Focus: Python development

Skills

Reusable capabilities that agents can invoke. Skills package specific workflows and can include templates, scripts, and references.

Golden Format Structure

skills/code-review/
├── SKILL.md
├── assets/       # Templates
├── scripts/      # Automation
└── references/   # Docs
💬

Commands

Slash commands that users invoke directly. They trigger specific workflows and can chain multiple skills together.

Usage Example

# In Claude Code terminal
/code-review src/main.py

# Invokes the code-review skill
# with the specified file

Plugin Structure

Standard directory layout for all plugins

custom-plugin-{role}/
custom-plugin-{role}/
├── .claude-plugin/
│   └── plugin.json          # Plugin manifest (required)
│
├── agents/
│   ├── {role}-expert.md        # Primary agent
│   ├── {role}-architect.md     # Architecture specialist
│   └── {role}-debugger.md      # Debugging specialist
│
├── skills/
│   └── {role}-skill/
│       ├── SKILL.md            # Skill definition
│       ├── assets/             # Templates, configs
│       ├── scripts/            # Automation scripts
│       └── references/         # Documentation
│
├── commands/
│   └── {role}-command.md       # Slash command
│
├── hooks/
│   └── hooks.json              # Event hooks
│
├── README.md                   # Plugin documentation
└── LICENSE                     # Apache 2.0

Frequently Asked Questions

How do I install a plugin?

In your Claude Code terminal, run:

/plugin install pluginagentmarketplace/custom-plugin-python

Then restart Claude Code to load the new plugin.

Can I modify the agents?

Absolutely! All plugins are Apache 2.0 licensed. You can:

  • Fork the repository
  • Edit agent.md files to change behavior
  • Add new skills and commands
  • Customize for your specific needs

See our Customization Guide for details.

What is SASMP protocol?

SASMP (Skill-Agent-Skill-Mapping Protocol) v1.3.0 ensures quality through:

  • Agent-Skill Bonding: Every skill has a responsible agent
  • EQHM: Event Quality Health Monitoring
  • Golden Format: Standardized skill structure

This guarantees consistency and quality across all 69 plugins.

How do I invoke an agent?

Use the Task tool to spawn a specialized agent:

Task(subagent_type="python-expert", prompt="Review this code")

Or use a skill:

Skill("python-code-review", args="src/main.py")

Need Help?

Can't find what you're looking for? We're here to help.