plugin-structure

To structure a Claude Code plugin for auto-discovery, use a standard directory layout with separate folders for commands, agents, skills, and hooks, then add a plugin.json manifest to define frontmatter fields like name, description, and version.

The plugin manifest in Claude Code is used to align configuration and establish consistent component loading by defining frontmatter fields such as name, description, and version, alongside portable path references like ${CLAUDE_PLUGIN_ROOT}.

You can use portable path references in Claude Code plugin architecture by implementing the ${CLAUDE_PLUGIN_ROOT} variable in your plugin.json manifest, ensuring scripts, references, and assets load reliably across different environments.

Yes, Claude Code plugin architecture supports MCP integration. The plugin structure provides best practices for organizing components, naming conventions, and configuring MCP server integration within your distribution-ready plugin kit.

The best way to scaffold commands and hooks for a Claude Code plugin is to follow a structured blueprint that places these components in clear directory layouts, allowing Claude Code to auto-discover them alongside your SKILL.md file.

You need a SKILL.md file in your plugin structure when you want to provide a structured blueprint for Claude Code plugin architecture, ensuring your commands, agents, skills, and hooks are properly documented and distribution-ready.