Transforms loose prompts into ones Claude can actually execute well. The core insight here is verification: adding test cases, screenshots, or concrete success criteria dramatically improves output quality. It also handles the context problem by either transforming directly or spinning up parallel agents to gather intent, codebase details, and relevant patterns first. The five principles (verification, context, constraints, phasing, rich content) are practical and specific. Most useful when you know what you want but your first pass at asking for it was too vague or missing the hooks Claude needs to self-correct.
npx skills add https://github.com/skillcreatorai/ai-agent-skills --skill best-practices