Plugin Structure

Organize commands and agents using standardized directory structures within your plugin. Place them in default directories or configure custom paths via plugin.json, following kebab-case naming conventions. This structure enables automatic component discovery and maintains plugin maintainability.

plugin.json defines plugin metadata and component paths. Specify manifest configuration, custom directory paths for commands, agents, skills, and hooks, and ensure proper relative path references. This file controls how Claude Code discovers and loads your plugin's components.

Use conventional directory layouts for automatic component discovery. Place plugin.json at the root, organize commands, agents, skills, and hooks in default or custom directories using kebab-case names. This standardized approach ensures seamless integration and optimal performance.

Claude Code discovers components from default directories automatically and from custom paths you define in plugin.json. The plugin loader checks relative paths and applies naming conventions to locate and load commands, agents, skills, and hooks without manual registration.

Yes, kebab-case naming is required for plugin components. This naming convention ensures consistent component loading rules and clear organization across default directories and custom paths defined in your manifest.

Default paths use built-in conventions for automatic discovery. Custom paths are explicitly configured in plugin.json to point to alternative locations for commands, agents, skills, and hooks, enabling flexible project layouts while maintaining clear loading rules.