review

UI Operator - Review

When to use

Use after capture/compare when the user wants a concise review of UI improvements, regressions, UX risks, and missing coverage.

Output contract

Writes only:

artifacts/ui-operator/<slug>/review.md

Never modify screenshots, comparison data, reports, or application code.

Pipeline

1. Read evidence

   • Open comparison-data.json.
   • Inspect report.html and key screenshots.
   • Read capture plan notes and blockers.

2. Classify findings

   • confirmed-regression: before/after evidence shows worse visible behavior.
   • ux-risk: behavior works but may confuse or slow the target user.
   • intentional-change: visible change matches the task or design direction.
   • coverage-gap: a planned state was not captured.
   • blocked: server, auth, selector, data, or browser prevented evidence.

3. Write review

   • Include severity, status, affected route/screen/state/viewport/device, evidence paths, observed behavior, impact, and next action.
   • Keep product/design judgment separate from factual browser evidence.
   • Prefer short findings over broad design essays.

4. Record limits

   • State missing routes, missing viewports, missing modals, broken baseline, or inconsistent data.

Hallucination guardrails

• Do not write a finding without evidence path or blocker.
• Do not infer user impact beyond the target workflow.
• Do not call an intentional visual change a bug unless it breaks usability or consistency.
• Do not claim accessibility compliance unless accessibility tooling was run.

Done criteria

• review.md lists findings ordered by severity.
• Every finding has evidence paths or a blocker.
• Coverage gaps and limits are explicit.
• Review distinguishes facts from judgment.