MCP HubMCP Hub
Вернуться к навыкам

review-ux-ui

pjt222
Обновлено Yesterday
5 просмотров
17
2
17
Посмотреть на GitHub
Дизайнdesign

О программе

Этот навык проводит комплексные UX/UI-аудиты с использованием эвристик Нильсена, проверок доступности по стандарту WCAG 2.1 и анализа пользовательских сценариев. Он предназначен для предрелизной проверки юзабилити, оценки соответствия требованиям доступности и эвристической оценки интерфейсов. Разработчики могут использовать его для систематической оценки когнитивной нагрузки, удобства форм и совместимости с клавиатурой и скринридерами.

Быстрая установка

Claude Code

Рекомендуется
Основной
npx skills add pjt222/agent-almanac -a claude-code
Команда плагинаАльтернативный
/plugin add https://github.com/pjt222/agent-almanac
Git клонированиеАльтернативный
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/review-ux-ui

Скопируйте и вставьте эту команду в Claude Code для установки этого навыка

Документация

Review UX/UI

Judge UX and UI for usability, accessibility, effectiveness.

When Use

  • Usability review before release
  • Check WCAG 2.1 AA or AAA compliance
  • Judge user flows for efficiency, error prevention
  • Review form design for usability, conversion
  • Heuristic eval of existing interface
  • Assess cognitive load, info architecture

Inputs

  • Required: App to review (URL, prototype, source code)
  • Required: Target user description (roles, skill, context)
  • Optional: User research findings (interviews, surveys, analytics)
  • Optional: WCAG target (A, AA, AAA)
  • Optional: Specific flows or tasks
  • Optional: Assistive tech to test (screen reader, switch)

Steps

Step 1: Heuristic Eval (Nielsen 10)

Rate interface against each heuristic.

#HeuristicKey QuestionRating
1Visibility of system statusDoes the system always inform users about what is happening?
2Match between system and real worldDoes the system use familiar language and concepts?
3User control and freedomCan users easily undo, redo, or exit unwanted states?
4Consistency and standardsDo similar elements behave the same way throughout?
5Error preventionDoes the design prevent errors before they occur?
6Recognition rather than recallAre options, actions, and information visible or easily retrievable?
7Flexibility and efficiency of useAre there shortcuts for experienced users without confusing novices?
8Aesthetic and minimalist designDoes every element serve a purpose? Is there unnecessary clutter?
9Help users recognize, diagnose, and recover from errorsAre error messages clear, specific, and constructive?
10Help and documentationIs help available and easy to find when needed?

Rate severity per violation.

SeverityDescription
0No problem
1Cosmetic — fix if time
2Minor — low priority
3Major — important fix
4Catastrophic — must fix before release
## Heuristic Evaluation Findings
| # | Heuristic | Severity | Finding | Location |
|---|-----------|----------|---------|----------|
| 1 | System status | 3 | No loading indicator during data fetch — users click repeatedly | Dashboard page |
| 3 | User control | 2 | No undo for item deletion — only a confirmation dialog | Item list |
| 5 | Error prevention | 3 | Date field accepts invalid dates (Feb 30) | Booking form |
| 9 | Error recovery | 4 | Form submission error clears all fields | Registration |

Got: All 10 heuristics rated with findings and severity.

If fail: Time short? Focus on heuristics 1, 3, 5, 9 — biggest impact.

Step 2: Accessibility Audit (WCAG 2.1)

Perceivable

  • 1.1.1 Non-text content: All images have alt text (decorative alt="")
  • 1.3.1 Info and relationships: Semantic HTML (headings, lists, tables, landmarks)
  • 1.3.2 Meaningful sequence: DOM order matches visual
  • 1.4.1 Use of colour: Color not sole info carrier
  • 1.4.3 Contrast: Text contrast ≥ 4.5:1 (normal), ≥ 3:1 (large)
  • 1.4.4 Resize text: Text resizes to 200% without breaking
  • 1.4.11 Non-text contrast: UI components, graphics ≥ 3:1
  • 1.4.12 Text spacing: Works with line height 1.5x, letter 0.12em, word 0.16em

Operable

  • 2.1.1 Keyboard: All function works via keyboard
  • 2.1.2 No keyboard trap: Focus never trapped
  • 2.4.1 Skip links: Skip nav for keyboard users
  • 2.4.3 Focus order: Tab order logical, predictable
  • 2.4.7 Focus visible: Focus indicator clearly visible
  • 2.4.11 Focus not obscured: Focused element not hidden behind sticky/overlays
  • 2.5.5 Target size: Targets ≥ 24x24px (44x44px touch)

Understandable

  • 3.1.1 Language of page: lang attr set on <html>
  • 3.2.1 On focus: Focus no surprise changes
  • 3.2.2 On input: Input no surprise changes without warning
  • 3.3.1 Error identification: Errors clear in text
  • 3.3.2 Labels or instructions: Form inputs have visible labels
  • 3.3.3 Error suggestion: Error msgs suggest fix

Robust

  • 4.1.1 Parsing: HTML valid (no dup IDs, proper nesting)
  • 4.1.2 Name, role, value: Custom components have ARIA roles
  • 4.1.3 Status messages: Dynamic changes announced to screen readers

Got: WCAG 2.1 AA criteria checked pass/fail per criterion.

If fail: Use auto tools (axe-core, Lighthouse) for first scan, then manual for human-judgment criteria.

Step 3: Keyboard and Screen Reader Audit

Keyboard Navigation Test

With Tab, Shift+Tab, Enter, Space, Arrow, Escape only.

## Keyboard Navigation Audit
| Task | Completable? | Issues |
|------|-------------|--------|
| Navigate to main content | Yes — skip link works | None |
| Open dropdown menu | Yes | Arrow keys don't work within menu |
| Submit a form | Yes | Tab order skips the submit button |
| Close a modal | No | Escape doesn't close, no visible close button in tab order |
| Use date picker | No | Custom date picker not keyboard accessible |

Screen Reader Test

Test with NVDA (Win), VoiceOver (mac/iOS), TalkBack (Android).

## Screen Reader Audit
| Element | Announced As | Expected | Issue |
|---------|-------------|----------|-------|
| Logo link | "link, image" | "Home, link" | Missing alt text on logo |
| Search input | "edit, search" | "Search products, edit" | Missing label association |
| Nav menu | "navigation, main" | Correct | None |
| Error message | (not announced) | "Error: email is required" | Missing live region |
| Loading spinner | (not announced) | "Loading, please wait" | Missing aria-live or role="status" |

Got: Full task flows tested with keyboard and screen reader.

If fail: No screen reader? Inspect ARIA attrs, semantic HTML as proxy.

Step 4: Analyse User Flows

Map and judge key flows.

## User Flow: Complete a Purchase

### Steps
1. Browse products → 2. View product → 3. Add to cart → 4. View cart →
5. Enter shipping → 6. Enter payment → 7. Review order → 8. Confirm

### Assessment
| Step | Friction | Severity | Notes |
|------|---------|----------|-------|
| 1→2 | Low | - | Clear product cards |
| 2→3 | Medium | 2 | "Add to cart" button below the fold on mobile |
| 3→4 | Low | - | Cart icon updates with count |
| 4→5 | High | 3 | Must create account — no guest checkout |
| 5→6 | Low | - | Address autocomplete works well |
| 6→7 | Medium | 2 | Card number field doesn't auto-format |
| 7→8 | Low | - | Clear order summary |

### Flow Efficiency
- **Steps**: 8 (acceptable for e-commerce)
- **Required fields**: 14 (could reduce with address autocomplete + saved payment)
- **Decision points**: 2 (size selection, shipping method)
- **Potential drop-off points**: Step 4→5 (forced account creation)

Got: Critical flows mapped with friction points, ratings.

If fail: No analytics? Judge flows by task complexity, step count.

Step 5: Assess Cognitive Load

  • Info density: Right amount per screen?
  • Progressive disclosure: Complex info revealed gradual?
  • Chunking: Related items grouped visual (Gestalt)?
  • Recognition over recall: Users see options vs remember?
  • Consistent patterns: Similar tasks use similar interactions?
  • Decision fatigue: Too many choices at once? (Hick's law)
  • Working memory: Need remember info across steps?

Got: Cognitive load assessed with overload/underload areas named.

If fail: Hard to judge objective? Use "squint test" — squint at screen, check structure and hierarchy still apparent.

Step 6: Review Form Usability

For each form.

  • Labels: Every input has visible, associated label
  • Placeholder text: Examples only, not labels
  • Input types: Right HTML types (email, tel, number, date) for mobile
  • Validation timing: Errors on blur or submit, not every keystroke
  • Error messages: Specific ("Email must include @") not generic ("Invalid input")
  • Required fields: Marked clear (mark optional if most required)
  • Field grouping: Related fields grouped visual (name, address, payment)
  • Autocomplete: autocomplete attrs set for standard fields
  • Tab order: Logical, matches visual layout
  • Multi-step forms: Progress indicator shows current, total steps
  • Persistence: Form data preserved if user navigates away

Got: Each form checked against list with issues documented.

If fail: Many forms? Prioritize highest-traffic (registration, checkout, contact).

Step 7: Write UX/UI Review

## UX/UI Review Report

### Executive Summary
[2-3 sentences: overall usability, most critical issues, strongest aspects]

### Heuristic Evaluation Summary
| Heuristic | Severity | Key Finding |
|-----------|----------|-------------|
[Summary table from Step 1]

### Accessibility Compliance
- **Target**: WCAG 2.1 AA
- **Status**: [X of Y criteria pass]
- **Critical failures**: [List]

### User Flow Analysis
[Key friction points with severity and recommendations]

### Top 5 Improvements (Prioritised)
1. **[Issue]** — Severity: [N] — [Specific recommendation]
2. ...

### What Works Well
1. [Specific positive observation]
2. ...

Got: Review gives prioritized, actionable recs with severity.

If fail: Too many issues? Split into "must fix" (severity 3-4) and "should fix" (1-2).

Checks

  • All 10 Nielsen heuristics rated
  • WCAG 2.1 criteria checked (min: 1.1.1, 1.4.3, 2.1.1, 2.4.7, 3.3.1, 4.1.2)
  • Keyboard nav tested for key flows
  • Screen reader tested (or ARIA/semantic HTML as proxy)
  • At least one critical flow analyzed
  • Cognitive load assessed
  • Form usability evaluated
  • Findings prioritized by severity, actionable recs

Pitfalls

  • Confuse UX with visual design: UX = how it works; visual = how it looks. Beautiful UI can have bad UX. Eval both but distinguish.
  • Test only happy path: Error states, empty states, loading, edge cases — UX problems hide there.
  • Ignore real devices: Dev tools responsive = proxy. Real devices catch touch, performance, viewport issues.
  • Accessibility as afterthought: Late = expensive. Evaluate early, continuous.
  • Personal preference as feedback: "I would prefer..." not UX feedback. Cite heuristics, research, patterns.

See Also

  • review-web-design — visual review (layout, typography, color)
  • scaffold-nextjs-app — Next.js scaffolding
  • setup-tailwind-typescript — Tailwind CSS design system

GitHub репозиторий

pjt222/agent-almanac
Путь: i18n/caveman/skills/review-ux-ui
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Похожие навыки

executing-plans

Дизайн

Используйте навык executing-plans, когда у вас есть полный план реализации для выполнения контролируемыми партиями с контрольными точками проверки. Он загружает и критически анализирует план, затем выполняет задачи небольшими партиями (по умолчанию 3 задачи), сообщая о прогрессе между каждой партией для проверки архитектором. Это обеспечивает систематическую реализацию со встроенными контрольными точками проверки качества.

Просмотреть навык

requesting-code-review

Дизайн

Этот навык запускает суб-агента для ревью кода, который анализирует изменения в коде на соответствие требованиям перед дальнейшими действиями. Его следует использовать после завершения задач, реализации крупных функций или перед слиянием с основной веткой. Ревью помогает выявить проблемы на ранней стадии, сравнивая текущую реализацию с исходным планом.

Просмотреть навык

connect-mcp-server

Дизайн

Этот навык предоставляет разработчикам подробное руководство по подключению серверов MCP к Claude Code с использованием транспортов HTTP, stdio или SSE. Он охватывает установку, конфигурацию, аутентификацию и безопасность для интеграции внешних сервисов, таких как GitHub, Notion и пользовательские API. Используйте его при настройке интеграций MCP, конфигурации внешних инструментов или работе с Model Context Protocol от Claude.

Просмотреть навык

web-cli-teleport

Дизайн

Этот навык помогает разработчикам выбирать между веб-интерфейсом Claude Code и CLI на основе анализа задачи, а также обеспечивает бесшовное перемещение сессий между этими средами. Он оптимизирует рабочий процесс, управляя состоянием и контекстом сессии при переключении между веб-интерфейсом, CLI или мобильным приложением. Используйте его для сложных проектов, требующих различных инструментов на разных этапах работы.

Просмотреть навык