---
name: experience-review
description: Review a rendered website or interface against its product and identity brief, with evidence-based usability, visual hierarchy, and customer-journey findings. Use for experience audits, not authorship detection or automatic redesign.
---

# Experience review

Inspect the actual experience before judging it. Establish the audience, intended action, approved identity, release status, requested scope, and source revision. Record the skill revision and brief revision, using unknown when unavailable. A screenshot supports visual observations; it does not prove links, keyboard behavior, delivery, or responsiveness.

## Inspect the relevant journey

Use the available browser or rendering tools and follow navigation and direct-entry paths. Select only relevant references:

- [Publisher sites](references/publisher-sites.md): a portfolio with products at different release stages.
- [Product pages](references/product-pages.md): audience, value, availability, and conversion paths.
- [Interactive experiences](references/interactive-experiences.md): state, feedback, help, input, and return paths.

Compare with approved identity decisions and the project's voice brief. Review hierarchy, product differentiation, availability labels, action semantics, responsive layouts, and applicable accessibility behavior. Inspect desktop and narrow renders when available. Test focus, keyboard paths, labels, contrast, reduced motion, and error states only where tools and authorization permit; describe exactly what was measured or exercised. If rendering is unavailable, provide a source-only review and leave visual conclusions unverified.

Do not submit live forms, make purchases, alter settings, or publish content without task authorization. Use local fixtures for destructive or externally visible states. Never infer delivery from a submit button, a public release from a marketing page, or privacy compliance from the absence of an obvious analytics script.

## Report useful findings

For each actionable finding, provide the URL or component, source file when verified, observed evidence, consequence for the visitor, classification, proposed correction, and testable acceptance criterion. Classify findings as defect, strategic tradeoff, aesthetic preference, or hypothesis; do not disguise preference as a failure. Prioritize critical broken journeys or exposure as P0, important usability/clarity problems as P1, and optional refinements as P2. Do not manufacture findings or numerical quality scores.

State what works and should remain. Separate observed facts from inferred causes and proposed changes. Include coverage and untested states. A passing automated check is not accessibility certification; clicks are not demand or purchases; a polished forecast interface is not accuracy evidence. Do not invent uplift, customers, measurements, or test results.

Recommend the smallest correction supported by the findings. Prose revisions can use an available editorial skill and the project voice brief, but this review remains independently usable. Record affected revisions so later changes can invalidate the relevant findings without pretending that an old report covers new work.