Audit Rendered UI

Audit Rendered UI

Use after opening the app in a browser, viewing supplied screenshots, or capturing simulator screenshots.

Required Input

Inspect actual pixels. Prefer browser, screenshot, image, or simulator evidence over code guesses. If the app cannot run, say the audit is code-only and lower confidence.

Viewports

Check at least:

• mobile: 390x844
• tablet: 768x1024
• desktop: 1280x900

For native apps, use the closest available simulator screenshots.

Audit Passes

1. Implementation-detail leak pass.

   • Flag user-visible copy about APIs, providers, databases, architecture, mocks, demos, local state, scaffolding, functions, model internals, or setup.
   • Ask whether a polished production app would say this to a normal user.

2. Redundancy pass.

   • Flag repeated titles, body copy, CTAs, save/confirm actions, suggestions, filters, and labels that restate nearby text.
   • Flag eyebrow labels or pretitles (small all-caps text above a heading) when they add only branding, metadata, or redundant hierarchy; keep them when they provide unique task context or required accessible meaning.
   • Prefer one element doing the work of two.

3. Text and layout fit pass.

   • No clipped labels.
   • No overlapping controls.
   • No loading labels that resize buttons.
   • Long words fit or wrap cleanly.

4. Visual clutter pass.

   • Flag nested cards, too many panels, crowded button rows, unneeded badges, oversized helper text, and competing primary actions.
   • Prefer removing or consolidating before restyling.

5. Interaction and asset pass.

   • Primary flows work.
   • Disabled, loading, error, and destructive states are understandable.
   • Maps, images, icons, and canvas or 3D surfaces are nonblank.
   • Provider attribution is present when required, but not framed as task copy.

6. Reference-alignment pass.

   • Compare against named references for hierarchy, density, typography, color roles, and rhythm.
   • Do not copy stale reference labels or deferred features.

Output

Lead with findings. For each finding include visible evidence, problem, fix, and severity: blocker, high, medium, or low.

If the audit is for an implementation agent, include a short subtractive edit plan with the smallest code areas to change.

References

Read references/leak-dedup-checklist.md for leaky words, redundant patterns, and replacement examples.