1,000 Stars: My Material Design 3 Skill Hit a Milestone I’d Never Reached

My Material Design 3 skill for AI coding agents just passed 1,000 stars on GitHub — my first repo ever to do so. The story, what it does (30+ MD3 components, Compose-first,...

Celebratory card in Material Design 3 purple gradient reading star 1,000 — Material Design 3 Skill — github.com/hamen/material-3-skill

My Material Design 3 skill just crossed 1,000 stars on GitHub — and it’s the first repo of mine to ever get there. I wanted to mark the moment, so: a short story, a thank-you, and a look at what the thing actually does.

I shipped it in early April 2026 as a single SKILL.md — a curated reference that teaches AI coding agents to get Material Design 3 right. Roughly ten weeks later it’s at 1,000+ stars and 52 forks. For an indie dev, that number doesn’t feel abstract; every star is someone who looked at it and thought “yes, I need this.”

Why a Material Design 3 skill?

AI coding agents are great at writing code and surprisingly shaky at design systems. Ask one for a Material 3 screen and you’ll often get plausible-looking UI with the wrong tokens, off-spec components, the wrong elevation, or color roles that don’t map to MD3 at all. The model knows “Material-ish,” not Material 3.

The skill closes that gap. It’s a distilled, curated reference — assembled from Google’s official m3.material.io docs, the Android Developers Compose Material 3 references, and the Google I/O 2026 updates — that an agent loads and follows.

What it does

In practice the skill helps an agent:

• generate MD3-compliant UI with correct design tokens, components, theming, layout and accessibility;
• cover 30+ components, with Jetpack Compose as the primary target, Flutter secondary, and Web (@material/web) as a documented but limited path;
• run an MD3 compliance audit that scores an existing app across 10 categories (Compose/Kotlin, Flutter/Dart, or web/CSS);
• apply M3 Expressive plus the layout and component updates from Google I/O 2026.

It was built the way I think a lot of good tooling will be built from now on: collaboratively, by a human and an AI agent, distilling public sources into something an agent can actually act on.

Try it

It’s MIT-licensed and one command to install in Claude Code:

claude plugin install github:hamen/material-3-skill

It also works with other skill-aware assistants — the skill content is plain SKILL.md + references. Repo here: github.com/hamen/material-3-skill.

Thank you

To everyone who starred, forked, opened an issue, or just tried it on a real screen: thank you. A thousand stars is a milestone I’d genuinely never hit before, and it’s pushed me to keep the skill current — the latest release already folds in the Google I/O 2026 Material guidance. More to come.