Skip to content
Agentmart
Submit a tool

visual-verdict

Screenshot comparison QA for frontend development.

Skill by vibeeval

IndexedMIT★ 540
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/vibeeval-visual-verdict

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Screenshot comparison QA for frontend development. Takes a screenshot of the current implementation, scores it across multiple visual dimensions, and returns a structured PASS/REVISE/FAIL verdict with concrete fixes. Use when implementing UI from a design reference or verifying visual correctness.

Use cases: design, frontend, testing

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/visual-verdict

What's inside

Skill name
visual-verdict
Description
Screenshot comparison QA for frontend development. Takes a screenshot of the current implementation, scores it across multiple visual dimensions, and returns a structured PASS/REVISE/FAIL verdict with concrete fixes. Use when implementing UI from a design reference or verifying visual correctness.
License
MIT
Instructions
286 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Visual Verdict — Screenshot QA Skill

Pixel-imprecise human eyes miss visual bugs. This skill provides structured, scored visual review using screenshots, returning actionable verdicts rather than vague impressions.

## When to Activate

- After a frontend-dev agent implements a UI component or page
- When cloning an external website (pairs with clone-website skill)
- After CSS changes to verify no visual regressions
- When a designer provides a Figma export or reference screenshot
- Before marking any UI task as complete

## How It Works

```
1. Take screenshot of current implementation
2. Load reference (design mockup, Figma export, or previous version)
3. Score against 6 dimensions (0-100 each)
4. Compute weighted total score
5. Issue verdict: PASS (90+) / REVISE (60-89) / FAIL (<60)
6. List concrete mismatches with file + line fix hints
7. Loop: frontend-dev fixes → new screenshot → rescore → repeat until PASS
```

## Prerequisites

- browser-use MCP installed and active (`~/.mcp.json`)
- Reference image available (Figma PNG export, screenshot, or URL)
- Implementation running (local dev server or deployed URL)

## Scoring Dimensions

### Dimension Weights

| Dimension | Weight | Description |
|-----------|--------|-------------|
| Layout accuracy | 0.25 | Positioning, spacing, alignment of all elements |
| Typography | 0.15 | Font family, size, weight, line-height, letter-spacing |
| Color accuracy | 0.15 | Background, text, border, shadow, gradient colors |
| Responsive behavior | 0.15 | Breakpoint transitions, mobile/tablet rendering |
| Interactive states | 0.15 | Hover, focus, active, disabled, loading states |
| Content completeness | 0.15 | All text, images, icons, labels present |

### Weighted Score Formula

```
total = (layout * 0.25) + (typography * 0.15) + (color * 0.15)
      + (responsive * 0.15) + (interactive * 0.15) + (content * 0.15)
```

### Scoring Each Dimension (0-100)

**Layout accuracy**
- All elements exactly in position, spacing matches: 90-100
- Minor spacing deviation (< 4px): 75-89
- Noticeable misalignment (4-16px): 50-74
- Layout clearly wrong (columns swapped, elements missing): 0-49

**Typography**
- Font, size, weight, line-height all match: 90-100
- One attribute off (wrong weight only): 70-89
- Two attributes off: 50-69
- Wrong font family entirely: 0-49

**Color accuracy**
- All colors within 5% of reference: 90-100
- One element color clearly off: 65-89
- Multiple elements wrong color: 40-64
- Dark/light mode inverted: 0-39

**Responsive behavior**
- Tested at 3 breakpoints, all correct: 90-100
- One breakpoint breaks layout: 65-89
- Two breakpoints break: 40-64
- Not responsive at all: 0-39

**Interactive states**
- All states implemented and visible: 90-100
- Missing one minor state (loading): 70-89
- Missing hover or focus: 50-69
- No interactive states at all: 0-49

**Content completeness**
- All text, images, icons present: 90-100
- One non-critical item missing: 75-89
- One critical item missing (CTA, heading): 50-74
- Multiple items missing: 0-49

## Verdict Thresholds

| Verdict | Score | Meaning |
|---------|-------|---------|
| PASS | 90-100 | Acceptable for production |
| REVISE | 60-89 | Needs fixes, not shippable |
| FAIL | 0-59 | Major rework required |

## Verdict Output Format

```markdown
## Visual Verdict: [Component/Page Name]

**Score: 85/100 — REVISE**
**Attempt: 2/5**
**Previous score: 72 (+13 this iteration)**

### Dimension Scores
| Dimension | Score | Weight | Weighted | Issues |
|-----------|-------|--------|----------|--------|
| Layout | 92 | 0.25 | 23.0 | Minor padding on mobile |
| Typography | 68 | 0.15 | 10.2 | Wrong font-weight on H2 |
| Color | 96 | 0.15 | 14.4 | - |
| Responsive | 80 | 0.15 | 12.0 | Card grid breaks at 768px |
| Interactive | 82 | 0.15 | 12.3 | Missing hover on CTA |
| Content | 91 | 0.15 | 13.7 | - |
| **TOTAL** | | | **85.6** | |

### Concrete Mismatches (prioritized by impact)

1. **[Typography / HIGH]** H2 hero heading

From vibeeval/vibecosystem, MIT.

More from vibeeval/vibecosystem

Signals

Stars
540
Forks
44
Open issues
5
Last commit (this folder)
Mar 28, 2026 (7 months ago)
Last release
Apr 12, 2026 (6 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from vibeeval

Cite this page

visual-verdict skill. Agent Mart. https://agentmart.dev/skills/vibeeval-visual-verdict (last updated 2026-10-24)