Compare Figma designs with rendered UIs using Playwright. Visual regression and structured reports for CI and experimental AI repair loops.

Two composable Claude Code skills for design-to-code fidelity: extract exact specs from Figma, then verify the running web or mobile app matches them.

Design DNA Phase 3 verification harness — scores fidelity (tokens, style, effects) and drives a correction loop until ship-ready.

Spacing & grid QA for AI-generated UIs, in the agent loop. MCP server that audits computed geometry and returns fix hints.

Design-fidelity QA for DeepSeek Harness: lend any text-only model an eye, then judge whether the implementation matches the mock. Ships the benchmark behind that judgement — four fixtures, 23 injected defects, and every raw model transcript. Retires itself when DeepSeek ships vision.

Visual QA CLI for AI-built frontends that renders pages, finds design and accessibility issues, and writes agent-ready fix prompts.

Stress-test Figma layouts for localization readiness, text expansion, RTL, and locale-specific formats.

Audit Figma flows for missing states, broken transitions, unsafe recovery paths, and unresolved product decisions.

Measure how closely a live page matches a design

Evidence-first UI alignment audit skill for AI coding agents (Claude Code / Cursor / Codex): real rendered evidence, typed findings, interactive A/B acceptance — without touching product code.

Design QA for the AI era — catch drift between your Figma designs and shipped UI. Pixel + element-level diff, auto-matching, token drift detection.

DESIGN-LAB（设计实验室）：面向职业视觉设计的、AI 原生、平台中立的设计智能与生产能力实验室。设计研究、合规知识、设计方法、领域能力、视觉质量、工具适配、生产预检、可编辑交付与证据体系，组织为可组合、可执行、可验证、可回滚的设计能力闭环。Host-native，当前参考宿主 Open Design（无默认绑定）。Agent-platform-neutral design intelligence & production laboratory for professional visual design.

Pre-ship design QA gate - contrast, touch targets, keyboard paths, breakpoints for any screen before it merges

Vibe coding is 30% building, 70% getting it to shipping quality. An agent skill that systematizes the 70% — a designer's pre-launch review for Claude Code, Cursor, or Codex that hands back a ranked, paste-ready fix list.

Portable Markdown skill for AI vibe coding with design QA: accessibility, WCAG, responsive testing, visual consistency, UI polish, and remediation plans.

Design QA Agent helps teams catch design-system drift in coded UI before it ships.

Visual UI regression agent for React: renders components with Playwright and uses Llama 4 Maverick (multimodal) to find visual defects and propose a unified-diff fix.

Structured style-fidelity checker for design-to-code: diff Figma vs. front-end per-attribute + by geometric boundary distance. AI judges only what needs judgment — a faster, cheaper, quantifiable alternative to screenshot visual diff.

Claude Code skill: pre-handoff state-coverage QA gate for ecommerce design files (PDP, cart, checkout)

MCP server + CLI for AI-powered design fidelity — compare Figma designs or screenshots against rendered UI. 90% mutation recall, zero false positives on a golden eval set. No API keys — uses your Claude Code or Codex subscription.