claude-code-plugin-dev

julep-ai's avatarfrom julep-ai

Comprehensive guide for building, testing, and distributing Claude Code plugins including slash commands, Agent Skills, subagents, MCP servers, and hooks. Use this when the user asks about creating plugins, writing slash commands, implementing skills, building MCP tools, configuring hooks, plugin architecture, marketplace distribution, or debugging plugin components. Covers 2025 schema with tool permissions, version tracking, and activation triggers.

⭐2stars🔀0forks📁View on GitHub🕐Updated Nov 24, 2025
[Developer Tools]

When & Why to Use This Skill

This Claude skill serves as a comprehensive technical guide for developers to build, test, and distribute Claude Code plugins. It covers the latest 2025 schema, providing expert knowledge on implementing slash commands, Agent Skills, MCP servers, and event-driven hooks to extend the agent's core capabilities and automate complex development workflows.

Use Cases

  • Creating custom slash commands to automate repetitive terminal tasks, git operations, or project scaffolding.
  • Developing auto-activating Agent Skills that provide specialized domain expertise triggered by specific file types or user intents.
  • Building Model Context Protocol (MCP) servers to integrate external data sources, proprietary APIs, and custom tools directly into the Claude environment.
  • Implementing post-tool-use hooks to automate code formatting, linting, or security scanning after Claude modifies files.
  • Architecting and packaging complex plugins for distribution on the Claude Code marketplace, including manifest configuration and version tracking.
nameclaude-code-plugin-dev
description"Comprehensive guide for building, testing, and distributing Claude Code plugins including slash commands, Agent Skills, subagents, MCP servers, and hooks. Use this when the user asks about creating plugins, writing slash commands, implementing skills, building MCP tools, configuring hooks, plugin architecture, marketplace distribution, or debugging plugin components. Covers 2025 schema with tool permissions, version tracking, and activation triggers."
version"1.0.0"
tags["plugin", "development", "skills", "commands", "mcp", "hooks", "agents", "marketplace", "tools", "claude-code", "sdk"]
allowed-tools"Read(*), Write(*), Edit(*), Bash(mkdir:*, git:*, npm:*, node:*, python:*, jq:*), Glob(**/*.md, **/*.json, **/*.ts, **/*.py)"

Claude Code Plugin Development Skill

Expert knowledge for building Claude Code plugins with up-to-date 2025 standards.

When Claude Should Use This Skill

Activate this skill automatically when the user:

  • Asks to create/build a Claude Code plugin
  • Wants to write slash commands or Agent Skills
  • Needs to implement MCP servers or custom tools
  • Wants to configure hooks or event handlers
  • Asks about plugin architecture or structure
  • Needs help with plugin.json manifest
  • Wants to publish to a marketplace
  • Asks about debugging plugin components
  • Mentions keywords: plugin, skill, command, MCP, hook, marketplace, subagent

Quick Overview

Claude Code plugins consist of five component types that extend functionality:

  1. Slash Commands - User-triggered shortcuts (/command)
  2. Agent Skills - Model-invoked capabilities (auto-activated)
  3. Subagents - Specialized task handlers
  4. MCP Servers - External tool/data integrations
  5. Hooks - Event-driven automations

Documentation Structure

This skill is organized into focused modules for easy reference:

Core References

  • Plugin Structure - Architecture, file organization, component types, scope hierarchy
  • Plugin Manifest - plugin.json schema, validation, configuration
  • Slash Commands - Command creation, arguments, tool permissions, namespacing
  • Agent Skills - 2025 schema, activation patterns, descriptions that work
  • MCP Servers - Custom tools, TypeScript/Python implementations, tool naming
  • Hooks - Event handlers, available events, matchers, environment variables

Practical Guides

Examples & Troubleshooting

Getting Started

For First-Time Plugin Creators

  1. Read Plugin Structure to understand architecture
  2. Create Plugin Manifest (required .claude-plugin/plugin.json)
  3. Choose component type(s) to implement:
  4. Follow Development Workflow for local testing
  5. Use Publishing Guide to distribute

For Specific Tasks

"I want to create a slash command" → See Slash Commands

"How do I make a skill that auto-activates?" → See Agent Skills - especially the "Activation Patterns" section

"I need to build custom tools/MCP server" → See MCP Servers

"How do I auto-format code after edits?" → See Hooks - PostToolUse event

"My skill isn't activating" → See Troubleshooting Guide - "Skills Not Activating" section

"What are the best practices?" → See Best Practices

Quick Reference

Minimal Plugin Structure

my-plugin/
├── .claude-plugin/
│   └── plugin.json              # REQUIRED
├── commands/                    # Optional
│   └── my-command.md
├── skills/                      # Optional
│   └── my-skill/
│       └── SKILL.md
├── mcp/                         # Optional
│   └── server.ts
└── README.md

Essential Commands

# Local testing
/plugin marketplace add ~/.claude/marketplaces/local
/plugin install my-plugin

# Management
/plugin list
/plugin update my-plugin
/help | grep my-command

# Validation
jq . .claude-plugin/plugin.json

Key 2025 Updates

  • ✅ allowed-tools field for tool permissions in skills
  • ✅ Version tracking required
  • ✅ Enhanced activation triggers with specific keywords
  • ✅ 240+ community plugins available

When to Reference Each Module

User Question Read This
"How do I structure a plugin?" Plugin Structure
"What goes in plugin.json?" Plugin Manifest
"Create a command that..." Slash Commands
"Build a skill that activates when..." Agent Skills
"Implement custom tools/API integration" MCP Servers
"Auto-run something after tool use" Hooks
"How do I test locally?" Development Workflow
"How do I publish my plugin?" Publishing
"My plugin component isn't working" Troubleshooting
"Show me a complete example" Examples

Important Notes

  • 2025 Schema: All documentation follows the latest 2025 plugin schema
  • Activation Keywords: Skill descriptions must be specific with file types, actions, and tools
  • Tool Permissions: Use allowed-tools to restrict access appropriately
  • Testing First: Always test locally before publishing
  • No Secrets: Never commit API keys or credentials

Additional Resources

Next Steps: Based on what the user is trying to build, direct them to the appropriate reference guide above. For comprehensive understanding, recommend reading in this order: Plugin Structure → Plugin Manifest → specific component type → Development Workflow → Publishing.