Back to directory

UX/UI Agent Skills

A kit of instructions, design tokens and runnable skills that turns an AI coding agent into a design-system generator, checked against dozens of automated accessibility and consistency gates.

UX/UI Agent Skills

UX/UI Agent Skills is a kit of structured instructions, design tokens and runnable skills, bundled with over a hundred reference design systems, that guides an AI coding agent to produce accessible, framework-agnostic components and pages instead of generic-looking output. It ships no build tools or runtime of its own; it is purely an instruction and knowledge layer that the agent reads.

Example outputs in the repository, such as a data-dashboard with several inline SVG chart types and a showcase page built only from the kit's tokens and rules, are rendered from the same files that dozens of objective gates measure, so the gates fail the build rather than accept output that only looks right. The project contrasts this against what an agent produces with no such rules: a generic gradient background, equal-sized cards with no focal point, emoji used as icons, and a single radius and shadow used everywhere.

It is distributed as a JavaScript-based package with no runtime dependency, installable as a Claude Code style skill or plugin.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.