AI coding agents are fast, but they cut corners. They skip specs, skip tests, skip security reviews. Agent Skills, by Addy Osmani, is a collection of 19 structured workflows designed to enforce proper engineering discipline at every phase of development.
The skills are organized across the full lifecycle: define, plan, build, verify, review, and ship. Each one includes step-by-step workflows with verification gates, so the agent can't just hand-wave past the hard parts. There are also anti-rationalization tables that push back on common excuses for skipping best practices, which is a practical addition.
The toolkit also includes 6 slash commands mapping to development phases (/spec, /plan, /build, /test, /review, /ship), 3 specialist agent personas (code-reviewer, test-engineer, security-auditor), and 4 reference checklists covering testing, security, performance, and accessibility.
It works with Claude Code, Cursor, Gemini CLI, Windsurf, and GitHub Copilot. MIT licensed with 9k+ GitHub stars.