agents-best-practices

Audit and improve existing agent skills against the agentskills.io specification and current best practices. Also audits CLAUDE.md, GEMINI.md, and AGENTS.md for bloat and configuration issues. Guides migration from client-specific folders to .agents/skills/ structure. Use when reviewing skill quality, checking specification compliance, optimizing descriptions for better triggering, identifying anti-patterns, upgrading skills to follow latest standards, or auditing agent instruction files. Activates on: "audit skills", "review my skill", "improve skill", "check skill quality", "skill audit", "fix skill description", "skill compliance", "check CLAUDE.md", "audit AGENTS.md", even if the user does not explicitly mention "audit" or "best practices". For creating new skills from scratch or running trigger evals, use skill-creator instead — this skill only audits and improves existing skills.

Finds existing agent skills before creating a new one. Searches a curated list of 1184+ official skills from Anthropic, Vercel, Stripe, Supabase, Cloudflare, Sentry, MongoDB, Figma, and 40+ other teams. Use when the user asks "find me a skill for X", "is there an existing skill for Y", "how do I do Z", or expresses a capability gap — even if they do not explicitly mention "skill". Run BEFORE skill-creator to avoid duplicating work.

Creates new agent skills and iteratively improves existing ones following the .agents/skills/ specification. Guides through intent capture, SKILL.md writing, testing, and description optimization. Use when creating a skill from scratch, turning a workflow into a reusable skill, improving an existing skill's content or triggering, or when the user says "make this a skill", "create a skill for X", "turn this into a skill", even if they don't explicitly mention "skill" but describe a repeatable workflow they want to automate.