uxskill
Star on GitHub
v4.0.1 · FOUNDATIONS · LIVE MIT · NO TELEMETRY · NO LLM

One brand color in. A design system out.

Give uxskill your brand color and a short brief. It builds a design system in nine foundations, checks every color pairing against WCAG in light, dark and high contrast, and writes it only when they all pass. Arabic and right to left built in.

Scroll↓
New in 4.0

Everything a product needs to look like one brand.

From one color and a short brief it works out the nine foundations below, measures every color pairing, and hands you the files.

A woman with long hair paddles a canoe on a still turquoise lake at dusk, a dark evergreen forest along the far shore.
Imagery is a foundation too. Each system names the photographs its pages use (light, temperature, grain, framing) and locks them to one grade.
  1. ColorScales and roles from your brand color, checked in every mode.
  2. TypeDisplay, text, mono and an Arabic face, sized for each script.
  3. SpacingOne scale for gaps and padding, comfortable or compact.
  4. LayoutContainer, margins, reading width and the gap between sections.
  5. RadiusCorners from sharp to round, one value per kind of part.
  6. BorderLine widths, control edges and a focus ring you can see.
  7. ElevationShadows and surfaces that keep depth readable in dark mode.
  8. MotionTiming for hover, press and arrival, calm under reduced motion.
  9. ImageryA photo direction every picture on the page shares.
tokens.csstokens.jsonfonts.cssart/system-report.md

Checked before it is written

Text on its background must reach 4.5:1 (WCAG 1.4.3), control edges and focus rings 3:1 (1.4.11), and text in high contrast 7:1 (1.4.6). If one pairing falls short in light, dark or high contrast, nothing is written and the report names it.

Arabic and right to left, built in

Every system carries an Arabic face and its own type scale. Set dir="rtl" and the text switches to them. There is no second system to keep in step.

Then the linter checks the page

171 rules flag what reads as generated and judge values against your own tokens. With --render it measures the real page in headless Chromium.

Already on 3.x? Nothing is rewritten for you. Read the guide to moving from 3.x to 4.0: build a new system and move the page onto its roles, or keep yours and extend it.

The thing you ship every day

Your AI ships the centroid by default.

Ask any model to "build a landing." You get the average of every page it ever ingested, in ten seconds, every time. Here is what the centroid looks like.

Build me a fintech landing page.
the slop
Right. I want it to look like us, not everyone.
That is the whole job. Read on. the turn ↓
The turn

What if the model had a designer’s brain wired in?

Not a bigger model. A deterministic layer that compiles a design language (warmth, contrast, geometry, type) instead of reaching for the nearest template.

One sequence, not scattered cards

It compiles a design language. Not a template.

Five deterministic steps, start to ship.

system.tokens.css
/* reading brief... */
--industry:'fintech'
--audience:'design-engineers'
--mood:'precise, calm'
01

Read the brief.

Industry, audience, mood, parsed into a structured intake, no guesswork.

system.tokens.css
/* synthesizing axes... */
--warmth:0.28
--contrast:0.91
--density:0.44
--formality:0.78
02

Set seven axes.

Warmth, contrast, density, geometry, formality, motion, type, tuned, then emitted as tokens.

system.tokens.css
/* anchoring brand... */
--color-surface-page:#f6f7f9
--color-text-default:#10131a
--color-action-primary:#5b9cff
--radius-card:10px
03

Anchor to your brand.

Primary pulled from your logo’s pixels, not the nearest CSS named colour.

system.tokens.css
--type-face-display:'Bricolage Grotesque'
--type-face-text:'Inter'
/* linting 171 rules... */
/* 0 violations */
04

Lint 171 anti-patterns.

Gradient headers, three equal cards, filler verbs, fails the build here.

system.tokens.css
/* verifying 360px... */
/* h-scroll: none */
/* sticky: 56px ok */
/* verified ok */
05

Verify mobile, for real.

Rendered at true 360/390px in headless Chrome. Horizontal scroll fails, not your users.

The corpus is vocabulary

160 real brand specs, as words, not stamps.

The catalogue used to be a menu: pick Stripe, get Stripe-flavoured tokens. v3 turned that on its head, Stripe, Linear, Vercel, Ferrari, Anthropic become the vocabulary the engine distils from, so output is novel every call. Deterministic, offline, no model in the loop.

1,262  structured entries, queryable and cross-referenced

corpus.manifest styles84 palettes176 type-pairs70 components148 brands160 anti-patterns171 1,262 entries deterministic · offline · no LLM
Runs where you already build

The same engine, the same brand-true output, everywhere.

Claude Code, Cursor, Windsurf, Copilot, Gemini, Codex, Zed, JetBrains, Cline, Continue, Aider, Tabnine, one install, no per-tool config. Your team keeps its editor; the floor comes with it.

Why this is the floor

Three properties. No competitor has all three.

Deterministic

Same brief, same system. No model in the engine, no telemetry, runs on a plane.

Brand-true

Primary from your logo's pixels; a fidelity hard floor fails any off-brand output.

Mobile-gated

Rendered at real 360 / 390px viewports, horizontal scroll fails the build, not your users.

1,262
Structured entries
160
Brand specs
171
Anti-pattern rules
17
IDE integrations
Install

One line. Then build.

Python$ pip install --upgrade uxskill
MCP$ pip install --upgrade 'uxskill[mcp]'
Node$ npx uxskill@latest