Agent Development

If you're building Claude Code plugins and need to create autonomous subprocesses that handle multi-step tasks without user intervention, this is the reference you want. It covers the core mechanics: structuring agents as markdown files with YAML frontmatter, writing triggering conditions with examples, designing system prompts that actually work, and customizing model behavior. The key insight here is understanding when to use agents versus commands, which comes down to autonomous execution versus user-initiated actions. It's documentation rather than a plugin itself, so treat it as a guide for building your own agent-based workflows in Claude Code.