# UI Skills

> A collection of design-engineering skills (prompt files) for humans and AI agents to create better UI interfaces, accessible via CLI or web browser.

UI Skills is an open-source collection of design-engineering skill files created by Julien Thibeaut (ibelick) and published under the MIT license. It provides curated prompt-based skills that help both humans and AI agents produce higher-quality, more polished user interfaces. The project is available on GitHub with nearly 5,000 stars and is actively maintained.

## What It Is

UI Skills is a registry of structured prompt files — called "skills" — that encode design-engineering expertise for use by AI coding agents. Each skill targets a specific UI concern: motion design, React performance, shadcn/ui component workflows, accessibility, animation principles, and more. Rather than replacing the agent, skills route it toward the right design knowledge before it makes changes. The collection is browsable on the web and installable via a CLI tool (`npx ui-skills`).

## How the Skill Routing Works

The core workflow is agent-first: a developer instructs their agent to run `npx ui-skills start`, which routes the agent through the appropriate skill for the task at hand. Skills are read-only advisors — they produce prioritized audit plans and implementation guidance for another model or agent to execute, rather than directly modifying source code. This separation keeps the skill layer focused on design intelligence and the execution layer focused on code changes.

## The Skill Collection

The collection spans a wide range of UI and design-engineering concerns, with skills contributed by notable community members and organizations:

- **ui-skills-root** — top-level router that selects the right skill by topic, stack, and intent
- **improve-ui** — audits existing product surfaces and writes implementation plans
- **frontend-design** (Anthropic) — generates creative, production-grade frontend code that avoids generic AI aesthetics
- **improve-animations** (emilkowalski) — motion advisor that audits animation code and produces improvement roadmaps
- **improve-react** (millionco) — React codebase auditor using React Doctor findings
- **shadcn** (shadcn-ui) — project-aware shadcn/ui workflow for components
- **12-principles-of-animation** — applies Disney's 12 animation principles to web interfaces
- **better-ui**, **design-taste-frontend**, **impeccable**, and more for visual polish and design taste

## CLI Interface

The `npx ui-skills` CLI is the primary access path for agent workflows. Key commands include:

- `npx ui-skills start` — routes the agent to the right skill
- `npx ui-skills categories` — lists available skill categories
- `npx ui-skills list --category motion` — filters skills by category
- `npx ui-skills get baseline-ui` — fetches a specific skill

## Update: v0.2.3

The latest release is v0.2.3, published in June 2026. The repository was last pushed in July 2026, indicating active development. The project was created in January 2026 and has accumulated nearly 5,000 GitHub stars and over 200 forks in a short period, signaling strong early traction in the design-engineering and AI agent tooling communities.

## Features
- CLI-based skill routing for AI agents
- Curated collection of design-engineering prompt skills
- Read-only audit and implementation planning skills
- Motion and animation improvement skills
- React codebase auditing
- shadcn/ui component workflow skill
- Accessibility and UX guideline review
- Disney 12 animation principles for web
- Frontend design taste and polish skills
- Community-contributed skills from notable developers
- Web browsable skill collection
- MIT licensed and open source

## Integrations
Claude / Anthropic agents, shadcn/ui, React, GSAP, npx / Node.js CLI

## Platforms
CLI, WEB, API

## Pricing
Open Source

## Version
v0.2.3

## Links
- Website: https://www.ui-skills.com/
- Documentation: https://www.ui-skills.com/skills
- Repository: https://github.com/ibelick/ui-skills
- EveryDev.ai: https://www.everydev.ai/tools/ui-skills