@sitinme: Google's Principal Engineer Wrote 24 AI Programming Skills, Packaging the Entire Development Lifecycle! Last week, I used Claude Code to modify an old API. The requirement was to add caching for two endpoints. I threw in a line: 'Help me add Redis caching' — it wrote it in three minutes, looked quite nice. …

X AI KOLs Timeline Tools
ai-programming open-source claude-code engineering-practices agent-skills

Summary

Google Principal Engineer Addy Osmani open-sourced agent-skills, which includes 24 AI programming skills, packaging engineering disciplines for use by AI agents to improve code quality and development efficiency.

Google's Principal Engineer wrote 24 AI programming Skills, packaging the entire development lifecycle! Last week, I used Claude Code to modify an old API. The requirement was to add caching for two endpoints. I first dropped a line: 'Help me add Redis caching' — it wrote it in three minutes, looked quite nice. But then running tests instantly turned red. The cache key was misspelled, expiration time unit was written in milliseconds, and worst of all, it casually refactored a completely unrelated piece of logic. I stared at the diff for over twenty minutes just to figure out what it had actually changed. This really can't be blamed on the model. It did what it was asked. The problem was that no one set the rules before it started: first clarify what's needed, break it into small tasks, handle them one by one, run tests on each piece, and only merge to the main branch when they pass. These steps are muscle memory for senior engineers when working on projects, but AI agents have no such muscle memory. Addy Osmani recently open-sourced something called agent-skills. In simple terms, it packages this engineering discipline and feeds it to AI. Addy is a Principal Engineer on the Google Chrome team and the author of 'Learning JavaScript Design Patterns'.
Original Article
View Cached Full Text

Cached at: 06/17/26, 11:54 AM

Google’s Chief Engineer Wrote 24 AI Programming Skills, Packaging the Entire Development Lifecycle!

Last week I used Claude Code to modify an old API; the requirement was to add caching to two endpoints.

I just dropped a line: “Help me add Redis caching.” It finished the code in three minutes, and it looked quite nice.

But when I ran the tests, everything turned red — the cache key was spelled wrong, the expiration time unit was in milliseconds, and most absurdly, it also casually refactored an entirely unrelated piece of logic. I stared at the diff for over twenty minutes just to figure out what it had actually changed.

You really can’t blame the model for this — it wrote what it was asked to. The problem is that no one set rules for it before it started: clarify what you want, break it into small tasks, tackle them one by one, run tests after each piece, and only merge into the main branch once they pass.

For seasoned engineers, these actions are muscle memory when working on projects, but AI agents simply don’t have that muscle memory.

Addy Osmani recently open-sourced a set of things called agent-skills — in short, it packages this engineering discipline into a format ready for AI consumption.

Addy is a chief engineer on the Google Chrome team and the author of Learning JavaScript Design Patterns.

Similar Articles

@gyro_ai: https://x.com/gyro_ai/status/2055198700016660826

X AI KOLs Timeline

Matt Pocock open-sourced Skills for Real Engineers, a set of small, composable, and hackable AI coding skills designed to address issues in AI programming such as understanding bias, lack of shared language, missing feedback loops, and software entropy. The tool enhances AI programming efficiency through skills like grill-with-docs, tdd, and diagnose, and provides a complete workflow.