The video introduces Addy Osmani's AI agent skills packages, designed to streamline senior engineering workflows. These packages function as plug-ins for AI coding assistants like Claude Code, Cursor, and Gemini CLI. The core concept is similar to Cursor rules but offers a more structured approach to managing the full lifecycle of skill files, complete with verification gates. The system comprises 20 skills mapped to seven commands: spec, plan, build, test, review, code simplify, and ship. It emphasizes spec-driven development, Test-Driven Development (TDD) principles like red-green-refactor, and the 80-15-5 test pyramid. Notably, the agent is capable of writing failing tests first, aligning with TDD practices. The video concludes by instructing viewers on how to install these skills via the Claude Code plug-in marketplace and encourages starring the GitHub repository if they utilize AI agents for shipping code.

VibeDev

Your AI agent skips the boring parts of senior engineering. Addy Osmani / agent skills packages senior engineering workflows as plug-in skills for Claude code, cursor, and Gemini CLI.

Think cursor rules, but structured as a full life cycle of skill files with verification gates baked in. 20 skills mapped to seven / commands: spec, plan, build, test, review, code simplify, ship.

Spec driven development, red green refactor TDD, the 80 15 5 test pyramid, anti-rationalization tables. Yes, your agent finally writes the failing test first.

In Claude code, run / plug-in marketplace at Addy Osmani / agent skills, then / plug-in install. Star the repo if you ship with AI agents.

agent-skills: Senior Engineering Workflows for AI Coding Agents

Your AI agent skips the boring parts of
senior engineering. Addy Osmani {slash}
agent skills packages senior engineering
workflows as plug-in skills for Claude
code, cursor, and Gemini CLI.
Think cursor rules, but structured as a
full life cycle of skill files with
verification gates baked in. 20 skills
mapped to seven {slash} commands: spec,
plan, build, test, review, code
simplify, ship. Spec driven development,
red green refactor TDD, the 80 15 5 test
pyramid, anti-rationalization tables.
Yes, your agent finally writes the
failing test first.
In Claude code, run {slash} plug-in
marketplace at Addy Osmani {slash} agent
skills, then {slash} plug-in install.
Star the repo if you ship with AI
agents.

Get Our Mobile App

agent-skills: Senior Engineering Workflows for AI Coding Agents

Transcription