| name | reviewing-frontend-aesthetics |
| description | Used by the Reviewer agent during Phase 9 (DesignVerify) and Phase 11 (Quality) to evaluate frontend code. Assesses visual hierarchy, spacing consistency, accessibility, and detects formulaic 'AI slop' design patterns, providing actionable feedback to the Developer. |
Reviewing Frontend Aesthetics (Phase 9 & 11: Reviewer)
Overview
As the Reviewer agent in Phase 9 (DesignVerify) and Phase 11 (Quality), your role is to perform a systematic, read-only quality check of the frontend code and generate a comprehensive audit report with prioritized issues and actionable recommendations.
CRITICAL: You are read-only regarding application source code (do not modify .ts, .css, etc.). Your only allowed write action is generating the audit report. You do not fix issues yourself. You document them systematically so the Developer (or other roles) can address them. Your goal is to catch all the small details that separate good work from great work, and to ruthlessly detect "AI slop" patterns.
Pre-Review Assessment
Understand the current state and goals:
- Review completeness: Is it functionally complete? Are there known issues to preserve (mark with TODOs)?
- Quality bar: MVP vs flagship feature?
- Design System: Read
AGENTS.md and the architecture plan to understand the intended aesthetic direction.
Diagnostic Scan (The Core Audit)
Conduct a holistic design critique across these dimensions. Think like a design director giving feedback.
1. AI Slop Detection (CRITICAL)
This is the most important check. Does this code/interface look like every other AI-generated interface from 2024-2025?
- Check for the AI color palette (cyan-on-dark, purple-to-blue gradients, neon accents).
- Check for identical card grids (icon + heading + text repeated endlessly).
- Check for generic hero copy ("Unlock the power of...", "Your all-in-one solution for...").
- Check for decorative blobs, floating circles, wavy SVG dividers.
- Check for centered everything (
text-align: center on all headings/cards).
- Check for uniform bubbly border-radius on every element.
- Check for default font stacks (Inter, Roboto, Arial, system).
The test: If you showed this to someone and said "AI made this," would they believe you immediately?
2. Visual Alignment & Spacing
- Pixel-perfect alignment: Everything lines up to grid.
- Consistent spacing: All gaps use a spacing scale (no random 13px gaps).
- Rhythm: Related items closer together, distinct sections further apart.
- Grid adherence: Elements snap to a baseline grid.
3. Typography & Consistency
- Hierarchy consistency: Same elements use same sizes/weights throughout. No skipped levels.
- Line length/height: 45-75 characters for body text. Appropriate line height.
- Font loading: No FOUT/FOIT flashes.
4. Color & Contrast
- Contrast ratios: All text meets WCAG AA standards (4.5:1 body, 3:1 large text).
- Consistent token usage: No hard-coded colors; all use design tokens/CSS variables.
- Tinted neutrals: No pure gray or pure black—subtle color tint required.
- Gray on color: Never put gray text on colored backgrounds.
5. Interaction States & Affordance
- Every interactive element needs all states: Default, Hover, Focus, Active, Disabled, Loading.
- Focus: Keyboard focus indicator (
focus-visible ring present, never outline: none without replacement).
- Touch targets: 44x44px minimum on touch devices.
6. Accessibility & Semantic HTML
- Missing ARIA: Interactive elements without proper roles, labels, or states.
- Keyboard navigation: Missing focus indicators, illogical tab order, keyboard traps.
- Semantic HTML: Improper heading hierarchy, missing landmarks,
divs instead of <button>s.
- Alt text: Missing or poor image descriptions.
- Forms: Inputs without labels, poor error messaging, missing required indicators.
7. Performance (Visual)
- Layout thrashing: Reading/writing layout properties in loops.
- Expensive animations: Animating layout properties (width, height, top, left) instead of transform/opacity.
- Images:
loading="lazy", width/height dimensions set, unoptimized assets.
Generate Comprehensive Report
Structure your feedback as a design director would. Write the report to the current phase's output directory (e.g., .jonggrang/.output/features/<feature-dir>/REVIEW_REPORT.md) using your Write tool.
Anti-Patterns Verdict
Start here. Pass/fail: Does this look AI-generated? List specific tells from the AI Slop section. Be brutally honest.
Executive Summary
- Total issues found (count by severity).
- Most critical issues (top 3-5).
- Recommended next steps.
Detailed Findings by Severity
For each issue, document:
- Location: Where the issue occurs (component, file, line).
- Severity: Critical / High / Medium / Low / Polish.
- Category: Accessibility / Performance / Theming / Responsive / Aesthetic / AI Slop.
- Description: What the issue is and why it matters (how it hurts users or undermines goals).
- Fix: What to do about it (be concrete: "Change X to Y because Z").
Critical Issues
[Issues that block core functionality or violate WCAG A]
High-Severity Issues
[Significant usability/accessibility impact, WCAG AA violations, egregious AI slop]
Medium-Severity Issues
[Quality issues, WCAG AAA violations, performance concerns, structural inconsistencies]
Polish Issues
[Minor inconsistencies, spacing tweaks, optimization opportunities]
Positive Findings (What's Working)
Highlight 2-3 things done well. Be specific about why they work to establish a baseline for quality.
Important Rules for Reviewing
- Be direct and specific. "The submit button padding" not "some elements". Vague feedback wastes everyone's time.
- Prioritize ruthlessly. If everything is important, nothing is.
- Don't soften criticism. Developers need honest feedback to ship great design.
- Never read source code for visual feedback if you can avoid it. If a headless browser command is available in your environment (e.g., via the bash tool), use it to evaluate the rendered site. Otherwise, evaluate the source code for systematic design consistency (tokens, semantic tags, anti-patterns).
- If the quality is terrible, say so. "This is a mess of hardcoded values and AI boilerplate."
Completion
Once the review report is generated and the findings are documented clearly, signal that your phase is complete.
Output the following completion signal exactly on its own line:
REVIEW_COMPLETE
