Addy Osmani's agent-skills repo offers 24 web-focused AI coding skills

This review examines Addy Osmani's agent-skills GitHub repository, detailing its 24 production-grade AI coding skills. It focuses on the unique emphasis on web performance and browser-based tasks for AI agents.

The Answer Up Front

For founders building AI coding agents that interact with web interfaces or require deep web performance optimization, addyosmani/agent-skills is a critical resource. It provides a comprehensive, opinionated set of skills covering the full software development lifecycle, with a unique focus on browser work and performance. Developers primarily focused on backend, data science, or non-web-centric tasks might find other skill frameworks like karpathy skills or superpowers more directly applicable. The bottom line is that this repository offers a robust, production-grade foundation for web-centric AI agents, leveraging the author's deep expertise.

Methodology

This v0 review draws on the founder's published claims as interpreted and detailed in a dev.to blog post, accessed on 2026-07-16. Independent benchmarks are pending. Update cadence: re-tested when claims diverge from observed behavior or when new versions are released. The tool under review is addyosmani/agent-skills, a GitHub repository of AI agent skills, version as of the dev.to post's observation. The source signal provides a detailed breakdown of the repository's contents, its author's background, and a comparative analysis with other popular skill repositories. This review covers the founder's stated intent, the enumerated skill list, and the architectural patterns described in the source. It does not cover independent performance benchmarks, long-term workflow integration, or edge-case behavior of agents utilizing these skills. Our assessment is based on the structural design and stated purpose of the skills, as no public artifacts for direct testing were provided in the source.

What It Does

addyosmani/agent-skills is a GitHub repository by Google web engineer Addy Osmani, containing 24 production-grade skills for AI coding agents. The repository, boasting 72.6k stars, distinguishes itself by embedding web performance and browser-based testing as first-class concerns, aligning with Osmani's specialization. It is presented as a plugin compatible with hosts like Claude Code, Cursor, and Copilot.

Full SDLC coverage

The skills are organized across the entire Software Development Lifecycle (SDLC): spec, plan, build, verify, review, and ship. This comprehensive approach means agents can be guided from initial idea refinement through to deployment and documentation. Examples include spec-driven-development for defining requirements, incremental-implementation for building, browser-testing-with-devtools for verification, and performance-optimization for review.

Web performance focus

A key differentiator is the explicit inclusion of skills tailored for web development, particularly performance. This includes browser-testing-with-devtools and performance-optimization. The source notes that this focus is unique among comparable skill repositories, making it particularly valuable for agents tasked with frontend development or optimizing web applications.

Excuse-preemption tables

Each skill incorporates an

The investor read

The market for AI agent skill frameworks is nascent but growing, driven by the need to make LLMs more reliable and effective in complex engineering tasks. addyosmani/agent-skills signals a trend towards specialized, domain-specific skill sets rather than generic prompts. While this is an open-source repository, its popularity (72.6k stars) and author's pedigree (Addy Osmani, Google) highlight the demand for high-quality, opinionated guidance for agent development. Comparable efforts include karpathy skills for general coding and superpowers for broader agentic workflows. An investable company in this space would likely offer a platform for managing, composing, and evaluating such skills, potentially with a marketplace for specialized, verified skill modules. This repo, while not a company, demonstrates the value of deeply integrating expert knowledge into agent capabilities, a critical component for future AI tooling spend.