Skip to content
Patterson Academy

Plugin structure and the manifest

A plugin bundles skills, commands, agents, hooks, and MCP servers into a single installable unit. Components are auto-discovered from standard directories:

my-plugin/ — standard layout
my-plugin/
  .claude-plugin/
    plugin.json        # manifest (the only required file)
  commands/            # slash commands (*.md)
  skills/              # skills (<name>/SKILL.md)
  agents/              # subagents (*.md)
  hooks/
    hooks.json         # hook configuration
  .mcp.json            # bundled MCP servers
  scripts/             # hook & utility scripts
  README.md
.claude-plugin/plugin.json
{
  "name": "plugin-dev",
  "description": "Comprehensive toolkit for developing Claude Code plugins",
  "version": "0.1.0",
  "author": {
    "name": "Patterson Engineering",
    "email": "eng@example.com"
  },
  "homepage": "https://github.com/example/plugin-dev",
  "license": "MIT",
  "keywords": ["plugins", "development", "hooks", "mcp"]
}

Plugin skills are namespaced (/my-plugin:review) so multiple plugins coexist. In hook commands and MCP configs, always reference files through ${CLAUDE_PLUGIN_ROOT} — the absolute path to the installed plugin — so the plugin works wherever it is installed. Test during development with claude --plugin-dir /path/to/my-plugin.

Check your understanding

Section titled “Check your understanding”

Question 1Which file is the only required one in a plugin?

Question 2Why use ${CLAUDE_PLUGIN_ROOT} in hook commands?

Question 3How are plugin skills invoked?