What problem does it solve? Self-built UI often ships with spacing drift, missing states, broken click-paths, Safari-only bugs, and slow Core Web Vitals that nobody measured. This Skill runs the full mechanical QA pass a skeptical engineer would run, producing a findings log of measured-vs-expected values instead of vague impressions. ## Core Features & Use Cases - Four ordered passes: fresh-eyes intent diff against the spec, automated click-path and state-consistency testing, a cross-device/cross-browser matrix, and a front-end performance pass covering Core Web Vitals, jank, layout thrash, and bundle size. - Scope decision tree: LIGHT for throwaway prototypes, STANDARD for internal tools, FULL for anything external-facing, so verification effort matches risk. - Neutral findings log: every row records expected value with source, measured value, and evidence, then hands off to craft-critique for severity and verdict. - Use Case: After building a portfolio case-study page, run a FULL verification to catch a 3.4s LCP caused by an unoptimized hero image, a missing focus-visible state on nav links, and a Safari backdrop-filter failure before a recruiter ever sees it. ## Quick Start Verify this page before ship: run the full pre-launch check on my production build and report every measured deviation from the design spec.
