This is a comprehensive template system for building autonomous Claude Code agents that handle complex, multi-step tasks in isolated contexts. It walks you through the YAML frontmatter structure (name, description, model, color, tools) and system prompt design, with heavy emphasis on the description field since that's what determines when Claude actually triggers your agent. The examples section is solid, covering four triggering patterns from explicit requests to proactive tool-based activation. What makes this useful is the specificity: you get concrete rules for naming conventions, a proven system prompt template with responsibilities and quality standards, and clear anti-patterns to avoid. If you're building agents beyond simple commands, this gives you the structure to make them trigger reliably and behave predictably.
npx skills add https://github.com/neolabhq/context-engineering-kit --skill create-agent