UI Design Review Skill

Overview

This skill provides prompting patterns, checklists, and templates for conducting UI design reviews using Gemini's multimodal vision capabilities.

When to Use

• Reviewing screenshots, wireframes, or mockups
• Conducting accessibility audits
• Validating design system consistency
• Comparing implementation to design reference
• Analyzing UI patterns and usability

Model Routing

Gemini API Key Detection

# Check API key availability and select model
determine_gemini_model() {
  if [[ -n "$GEMINI_API_KEY" ]]; then
    echo "g/gemini-3-pro-preview"  # Direct Gemini API
  elif [[ -n "$OPENROUTER_API_KEY" ]]; then
    echo "or/google/gemini-3-pro-preview"  # OpenRouter
  else
    echo "ERROR: No API key available (need GEMINI_API_KEY or OPENROUTER_API_KEY)"
    return 1
  fi
}

GEMINI_MODEL=$(determine_gemini_model)

Why or/ Prefix for OpenRouter

The model ID google/gemini-3-pro-preview collides with Claudish's automatic routing:

Rule: Always use or/ prefix when routing Google models through OpenRouter.

Passing Images to Claudish

Method 1: Image File Path (Recommended)

# Pass image file directly with --image flag
npx claudish --model "$GEMINI_MODEL" --image "$IMAGE_PATH" --quiet --auto-approve <<< "$ANALYSIS_PROMPT"

# Example
npx claudish --model "g/gemini-3-pro-preview" --image "screenshots/dashboard.png" --quiet --auto-approve <<< "Analyze this UI for usability issues."

Method 2: Base64 Encoded (For Inline Images)

# Encode image to base64
IMAGE_B64=$(base64 -i "$IMAGE_PATH")

# Include in prompt with data URI
printf '%s' "[Image: data:image/png;base64,$IMAGE_B64]

Analyze this UI for usability issues." | npx claudish --stdin --model "$GEMINI_MODEL" --quiet --auto-approve

Method 3: URL Reference (For Remote Images)

# Reference image by URL
printf '%s' "[Image: https://example.com/screenshot.png]

Analyze this UI for usability issues." | npx claudish --stdin --model "$GEMINI_MODEL" --quiet --auto-approve

Prompting Patterns

Pattern 1: General Usability Review

Analyze this UI screenshot for usability issues.

**Image**: [attached or path]

**Focus Areas**:
1. Visual hierarchy - Is the most important content prominent?
2. Affordances - Do interactive elements look clickable/tappable?
3. Feedback - Is system status clearly communicated?
4. Consistency - Do similar elements behave similarly?
5. Error prevention - Are destructive actions guarded?

**Output Format**:
For each issue found:
- **Location**: Where in the UI
- **Issue**: What the problem is
- **Principle**: Which design principle it violates
- **Severity**: CRITICAL/HIGH/MEDIUM/LOW
- **Recommendation**: Specific fix

Pattern 2: WCAG Accessibility Audit

Audit this UI for WCAG 2.1 AA compliance.

**Image**: [attached or path]

**Checklist**:
1. **Perceivable**
   - [ ] Text contrast >= 4.5:1 (WCAG 1.4.3)
   - [ ] Non-text contrast >= 3:1 (WCAG 1.4.11)
   - [ ] Information not conveyed by color alone (WCAG 1.4.1)
   - [ ] Text resizable to 200% (WCAG 1.4.4)

2. **Operable**
   - [ ] Keyboard accessible (WCAG 2.1.1)
   - [ ] No keyboard traps (WCAG 2.1.2)
   - [ ] Focus visible (WCAG 2.4.7)
   - [ ] Touch targets >= 44x44px (WCAG 2.5.5)

3. **Understandable**
   - [ ] Labels present for inputs (WCAG 3.3.2)
   - [ ] Error identification clear (WCAG 3.3.1)
   - [ ] Instructions available (WCAG 3.3.2)

4. **Robust**
   - [ ] Valid structure implied (headings, regions)

**Output Format**:
| Criterion | Status | Notes | Fix |
|-----------|--------|-------|-----|
| 1.4.3 | PASS/FAIL | Details | Recommendation |

Pattern 3: Design System Consistency Check

Compare this implementation against the design system.

**Implementation Image**: [attached or path]
**Design System Reference**: [tokens, components, patterns]

**Validation Points**:
1. **Colors**
   - Primary, secondary, accent colors
   - Semantic colors (success, warning, error)
   - Background and surface colors

2. **Typography**
   - Font family usage
   - Size scale adherence
   - Weight usage (regular, medium, bold)
   - Line height consistency

3. **Spacing**
   - Margin scale (4, 8, 16, 24, 32, 48...)
   - Padding consistency
   - Gap between elements

4. **Components**
   - Button variants (primary, secondary, ghost)
   - Input styles (default, focus, error, disabled)
   - Card patterns

5. **Elevation**
   - Shadow levels
   - Border usage
   - Layer hierarchy

**Output Format**:
| Element | Expected | Actual | Deviation |
|---------|----------|--------|-----------|
| Button BG | #2563EB | #3B82F6 | Wrong shade |

Pattern 4: Comparative Design Review

Compare the implementation screenshot to the original design.

**Design Reference**: [Figma export or mockup]
**Implementation**: [Screenshot of built UI]

**Comparison Points**:
1. Layout and positioning accuracy
2. Color fidelity
3. Typography matching
4. Spacing precision
5. Component rendering
6. Responsive behavior (if multiple sizes)

**Output Format**:
## Match Analysis

**Overall Fidelity**: X/10

### Exact Matches
- [List elements that match perfectly]

### Deviations
| Element | Design | Implementation | Impact | Fix |
|---------|--------|----------------|--------|-----|
| CTA Button | #2563EB | #3B82F6 | Visual | Change to design color |

### Missing Elements
- [Elements in design but not in implementation]

### Extra Elements
- [Elements in implementation but not in design]

Review Templates

Quick Review (5 minutes)

Focus on critical issues only:

1. Can users complete the primary task?
2. Are there any major accessibility barriers?
3. Is the visual hierarchy clear?
4. Are interactive elements obviously interactive?

Standard Review (15 minutes)

Full heuristic evaluation:

1. All Nielsen's 10 heuristics
2. WCAG AA key criteria
3. Visual design quality
4. Interaction design

Comprehensive Review (30+ minutes)

Deep analysis including:

1. Full heuristic evaluation
2. Complete WCAG AA audit
3. Design system consistency
4. Competitive analysis context
5. User flow mapping

Severity Guidelines

Integration with Multi-Model Validation

The ui agent supports PROXY_MODE for multi-model design reviews:

Task: ui PROXY_MODE: or/google/gemini-3-pro-preview

Review the dashboard screenshot at screenshots/dashboard.png

Focus on usability and accessibility.
Write review to: ${SESSION_PATH}/reviews/design-review/gemini.md

This enables:

• Running multiple design reviewers in parallel
• Consensus analysis on design issues
• Different model perspectives (Gemini vision, Claude reasoning)

Best Practices

DO

• Always validate image inputs exist before analysis
• Cite specific design principles for every issue
• Provide actionable, specific recommendations
• Prioritize by severity (CRITICAL first)
• Use Gemini for visual analysis (not guessing)

DON'T

• Give vague aesthetic opinions ("looks bad")
• Overwhelm with LOW severity items
• Forget accessibility considerations
• Skip the principle citation
• Assume implementation details without seeing