Agent Skills encodes senior engineering discipline into AI coding agent workflows
Agent Skills is an open-source collection of 24 production-grade engineering workflow files for AI coding agents, built by Addy Osmani of Google Chrome, designed to prevent agents from defaulting to shortcuts that skip tests, specs, and security reviews.
Score breakdown
Agent Skills directly addresses the accumulation of technical debt in AI-assisted development by replacing ad-hoc agent improvisation with structured, exit-criteria-driven workflows that enforce the engineering discipline agents skip by default.
- 01Created by Addy Osmani, Principal Engineer at Google Chrome and author of Learning JavaScript Design Patterns
- 02Addresses AI agent default behavior of skipping tests, specs, input validation, and security reviews in favor of the shortest path to working code
- 03Contains 24 skills covering 7 phases of the development lifecycle
Agent Skills, published under the MIT license by Addy Osmani — Principal Engineer on the Google Chrome team and author of *Learning JavaScript Design Patterns* — is a collection of production-grade engineering workflow files intended to serve as "the discipline layer your AI agent is missing." The project starts from the observation that AI coding agents are not incapable, but that their default behavior drives them toward the shortest path to working code, systematically deferring specs, test coverage, security hardening, and architecture documentation as non-essential to "runs."
When an agent processes a relevant task, it reads the skill file and follows the defined steps and checkpoints rather than improvising.
The project structures its solution across 24 skills that cover 7 phases of the development lifecycle, 7 slash commands forming a complete `/spec` to `/ship` workflow, and 4 specialized agent personas: Code Reviewer, Test Engineer, Security Auditor, and Web Performance Auditor. Each skill is defined in a `SKILL.md` file containing anti-rationalization tables and verification exit conditions, embedding engineering principles such as Hyrum's Law, the Beyoncé Rule, and Chesterton's Fence directly into the workflows. When an agent processes a relevant task, it reads the skill file and follows the defined steps and checkpoints rather than improvising. The repository has accumulated over 51,900 GitHub stars and 5,700 forks and is compatible with Claude Code, Cursor, and other AI coding tools.
Key facts
- 01Created by Addy Osmani, Principal Engineer at Google Chrome and author of Learning JavaScript Design Patterns
- 02Addresses AI agent default behavior of skipping tests, specs, input validation, and security reviews in favor of the shortest path to working code
- 03Contains 24 skills covering 7 phases of the development lifecycle
- 04Includes 7 slash commands forming a /spec to /ship workflow
- 05Features 4 agent personas: Code Reviewer, Test Engineer, Security Auditor, and Web Performance Auditor
- 06SKILL.md files contain anti-rationalization tables and verification exit conditions
- 07Repository has 51,900+ GitHub stars and 5,700+ forks; licensed MIT
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 11, 2026 · 08:34 UTC. How this works →