Subagent Creator

This is the meta-skill for building subagents, the specialized AI assistants that run in isolated contexts when your main agent needs to delegate complex work. You'd use it when you need a verifier to validate completed tasks, a debugger to hunt down root causes, or a security auditor to review sensitive code. The guide emphasizes that the description field is critical because it determines when automatic delegation happens, and it walks you through the frontmatter setup, prompt structure, and common patterns like code reviewers and validators. If you're building workflows where different specialized perspectives matter more than raw context accumulation, this shows you how to architect that separation properly.