Aetox Ux Review

---
name: aetox-ux-review
description: ตรวจและวิจารณ์ UX ของ UI ที่ทำไปแล้ว ด้วยกรอบมาตรฐาน, Nielsen 10 heuristics, Don Norman 7 หลัก, WCAG POUR, UX audit (IxDF 7 ปัจจัย), รีวิวงานภาพ และ cognitive walkthrough รายทาสก์ ใช้ตอนอยากประเมินของที่มีอยู่ ไม่ใช่ตอนสร้างใหม่ (สร้างใหม่ดู aetox-frontend-design)
source: https://github.com/mastepanoski/claude-skills
license: MIT
copyright: Copyright (c) 2026 mastepanoski. Full terms in LICENSE
---

# UX Review

Evaluating a UI that already exists, not designing one. Each method below is a
different lens; open the one the question calls for and read it whole before
judging. They are audits, so they end in findings, not opinions.

Pick the lens by what you are asking:

- **Usability, broadly**, `references/nielsen.md`. Jakob Nielsen's 10 heuristics:
  visibility of status, match to the real world, user control, consistency,
  error prevention, recognition over recall, flexibility, minimalist aesthetics,
  error recovery, help. The first reach for a general "is this usable" pass.
- **Why an interaction confuses**, `references/don-norman.md`. Don Norman's 7
  principles: discoverability, affordances, signifiers, feedback, mapping,
  constraints, conceptual model. Reach here when something is technically fine
  but people still do not understand it.
- **Accessibility**, `references/wcag-pour.md`. WCAG 2.1/2.2 across the four
  POUR principles at A/AA/AAA. The audit to run before calling anything done.
- **A whole experience, ending in a redesign**, `references/ux-audit.md`. IxDF's
  7 factors, 5 usability characteristics and 5 interaction dimensions, closing on
  redesign proposals rather than a bare list.
- **The visual layer**, `references/visual-review.md`. Typography, colour,
  spacing, hierarchy, consistency and branding against current category norms.
- **One task, step by step**, `references/cognitive-walkthrough.md`. Simulates a
  first-time user through a specific task to find where learnability breaks.

Run the narrowest lens that answers the question; reach for a second only when
the first surfaces something it cannot judge.

## What a finding is

Same bar whichever model is auditing:

- Each finding names the screen or step, the heuristic or criterion it
  fails, the evidence (a `browser` capture, or the exact element and text),
  who it hurts, and the fix. A finding without evidence is an opinion and is
  cut before the report.
- Ranked by severity, blockers first: cannot complete the task, cannot
  perceive or operate it, then friction, then polish. The count of findings
  is never the headline; the worst one is.
- WCAG 2.2 AA is the floor under every lens, not a lens you may skip. A
  visual review that praises the type on a page with 3:1 body text has
  reviewed the wrong thing first.
- The report ends with what to do first, in order, and what was not
  reviewed. A clean report on a partial audit says which parts were not
  looked at.

Adapted from mastepanoski's claude-skills (MIT), the six audits kept whole,
one file each.