UX-audit

Audit one page or surface at a time against the Designing Interfaces pattern language and the ux skill checklists.
Run static, visual, or dynamic review layers as needed and record findings with layer-appropriate evidence.
Report patterns in use, strengths worth preserving, and ranked UX gaps.

The skill reviews one surface per run; it does not perform a full-app sweep in a single pass.
Code review alone cannot establish visual or runtime conclusions.
Visual review cannot establish in-progress, locked, capability-gated, focus, keyboard, or quantified performance behavior.
Without funnel data, variant-winner conclusions have insufficient evidence.
Check empty, error, retry, and draft-persistence branches during static review.
Check rendered hierarchy, dominant control, spacing, contrast, alignment, truncation, overflow, and responsive breakpoints during visual review.
Check user-journey progression, in-progress and locked states, forced error or empty states, focus order, and keyboard reachability during dynamic review.
Capture CLS, LCP, INP, and long-task values only through dynamic review.

AI 搜索