
Description
Ask an AI to build a UI and nine times out of ten you get the same purple gradient and rounded cards, with contrast that fails, keyboard navigation that dead-ends and a layout that overflows sideways on a narrow screen. Someone always has to redo it.
ux-ui-agent-skills, or XIAS, is a design doctrine you feed to coding agents such as Claude, Codex and Cursor. It puts 138 brand-grade design systems, 52 component specs and 25 runnable skills into your project so the agent builds from real design systems and tokens instead of guessing.
What sets it apart is 52 objective gates, 34 of which open a real browser and measure: WCAG contrast in every state, keyboard operation, target size, reduced motion and whether anything overflows horizontally at 280px. A pass or a fail, not the AI grading itself. It is pure instructions and knowledge, with no build tools and no runtime dependencies.
Design system library: 138 design systems supply colors, type and spacing for the style you pick.
Component specs: 52 component specs describe states, interaction and accessibility, so you stop re-explaining them.
Design tokens: 14 DTCG token files, with colors and sizes chosen by intent rather than hard-coded.
Framework adapters: 16 adapters covering React, Next.js, SwiftUI, Tailwind and more.
Runnable skills: 25 skills for auditing, generating and fixing, callable right inside the coding agent.
Objective gates: 52 checks that measure accessibility and responsiveness in a real browser, with a plain pass or fail.
ux-ui-agent-skills, or XIAS, is a design doctrine you feed to coding agents such as Claude, Codex and Cursor. It puts 138 brand-grade design systems, 52 component specs and 25 runnable skills into your project so the agent builds from real design systems and tokens instead of guessing.
What sets it apart is 52 objective gates, 34 of which open a real browser and measure: WCAG contrast in every state, keyboard operation, target size, reduced motion and whether anything overflows horizontally at 280px. A pass or a fail, not the AI grading itself. It is pure instructions and knowledge, with no build tools and no runtime dependencies.
Features
Design system library: 138 design systems supply colors, type and spacing for the style you pick.
Component specs: 52 component specs describe states, interaction and accessibility, so you stop re-explaining them.
Design tokens: 14 DTCG token files, with colors and sizes chosen by intent rather than hard-coded.
Framework adapters: 16 adapters covering React, Next.js, SwiftUI, Tailwind and more.
Runnable skills: 25 skills for auditing, generating and fixing, callable right inside the coding agent.
Objective gates: 52 checks that measure accessibility and responsiveness in a real browser, with a plain pass or fail.
