Agentic Design Systems

Part 1 of the Agentic Design Systems series, based on Romina's IDS 2026 talk. A design system built for AI agents, not just humans. Stops hallucination before it ships.

Why AI coding tools need both a project rules file and a tools file, what belongs in each one, and how to connect them so agents stop guessing your stack.

Why AI-generated websites look the same, and how to fix it with a three-layer system that teaches your AI tool your visual standards, motion language, and copy tone.

The one protocol that lets AI read your Figma, GitHub, and docs directly. A plain-English explainer for designers, with examples of what it unlocks in real work.

Stop copy-pasting screenshots into AI. Set up Figma Dev Mode MCP so Claude and Cursor can read your designs directly, then ship code that actually matches your file.

End-to-end workflow for converting Figma designs to production code using MCP servers. From design to implementation in minutes, not hours.

Export Figma variables and transform them into production-ready CSS custom properties your engineers can use immediately.

Programmatically access your Figma components, styles, and variables. Build audits, reports, and automations on real data.

Why the next generation of design systems will be measured in skills, not components. A strategic essay on the shift from visual tokens to promptable taste.

A hands-on workshop to convert your personal design taste into a /taste-skill command. By the end, Claude will generate UI that looks like yours, not like everyone else's.

Build a working UX review agent in 30 minutes. Point it at any screen, get back a prioritized list of first-time user issues, accessibility problems, and copy fixes. The concrete 'aha' moment of Claude Code for designers.

A 5-version walkthrough showing how a bare prompt becomes a production-ready design audit. Built on the same iterative method Anthropic teaches internally.

Part 2 of the Agentic Design Systems series. The five trust levels (Observer, Advisor, Junior, Senior, Autonomous), what each one is allowed to do, and the exact criteria to promote an agent from one level to the next.

A step-by-step workflow to audit any design system using AI tools. Check token naming, component health, accessibility, and visual consistency in one session.

Mintlify is $120 a month. Supernova is enterprise pricing. Astro Starlight is free, fast, and good enough for most design system teams. Here is the full setup, the styling, the Figma embeds, and the AI-assisted automation.