Agent skill

Design System Generator Skill: Write a DESIGN_SYSTEM.md

The design-system-generator skill has an AI coding agent inspect your project, then write or revise a design system document covering tokens, components, accessibility, motion and visual checks.

Install
/plugin marketplace add thienanblog/awesome-ai-agent-skills
View on GitHub

Stars, license and last push from GitHub. Install count as reported by the open skills directory.

What is the design-system-generator skill?

Design-system-generator is a skill in the awesome-ai-agent-skills collection. Its job is a durable design system document grounded in your real stack. The agent reads existing docs, screens, components and tokens first, then writes the project contract, and verifies paths, token names and rendered output before reporting.

Design-system-generator ships inside the project-development-skills bundle of thienanblog's awesome-ai-agent-skills repository. It is for one situation: the durable design system document is the deliverable, not a one-off UI edit.

The description spells out the scope: tokens, components, accessibility, motion and visual verification. The agent is told to keep the output concise and tied to the product's actual stack and visual language.

What the agent does, in order

  1. Inspect first. It reads existing design docs, brand references, representative screens, shared components, theme configuration, tokens and build tooling, and infers framework, CSS approach, themes and export formats.
  2. Preserve what works. An established, coherent system is kept. For a new product, it recommends one direction within your constraints.
  3. Write the contract, using a bundled template as a menu of topics rather than a form to fill.
  4. Verify paths, token names, component APIs and build claims against their owners, then report.

What the document covers

  • Surfaces, themes, devices, and any language or right-to-left constraints.
  • Token ownership, typography, color, spacing, layout and motion.
  • Shared components and the interaction states each one really supports.
  • Extension rules for tokens, components, utilities and local exceptions.
  • Keyboard interaction, visible focus, contrast targets, semantic markup and reduced motion.
  • Visual verification at comparable states and viewports, plus asset loading and cache invalidation when the design system owns them.

Where the skill stops

  • It is not for a single screen. A one-off UI edit does not need a whole document.
  • It does not add a component library or migrate build tooling just to match its own examples.
  • It omits prompt-writing advice and generic framework tutorials. For component rules see shadcn, and for a Figma-side library see figma-generate-library.

Use it with UiChemy

A written design system gives your agent fixed tokens to build against. UiChemy then moves the finished build into WordPress: AI to WordPress imports an AI build, and Figma to WordPress converts a Figma design. Click Get Started on uichemy.com. For the colors, see color-mode-and-theme.