Claude Code Plugins

Community-maintained marketplace

Feedback

page-diff

@j-r-beckett/SpeedReader
0
0

Visual regression testing for web pages. Compares screenshots to detect UI changes, generates overlay images highlighting differences. Use after making frontend changes to verify visual correctness, catch unintended side effects, or validate that changes look as expected.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name page-diff
description Visual regression testing for web pages. Compares screenshots to detect UI changes, generates overlay images highlighting differences. Use after making frontend changes to verify visual correctness, catch unintended side effects, or validate that changes look as expected.

Page Diff

Compare screenshots to detect visual changes. Outputs similarity percentage and an overlay image with differences highlighted in magenta.

Usage

uv run .claude/skills/page-diff/scripts/diff.py <expected> <actual>

Output:

PASS: 99.2% similar (3,247 pixels differ)
/path/to/overlay.png

The overlay image shows the actual screenshot with changed regions tinted magenta. Use Read to view it.

Baselines

Store known-good screenshots in baselines/:

.claude/skills/page-diff/baselines/
├── demo.png
├── settings.png
└── upload-complete.png

Workflow:

  1. Capture baseline: browser_take_screenshot → save to baselines/demo.png
  2. Make changes, reload
  3. Capture current: browser_take_screenshot → /tmp/current.png
  4. Compare: uv run .claude/skills/page-diff/scripts/diff.py baselines/demo.png /tmp/current.png
  5. If diff is expected, update baseline

Interpreting Results

  • 100% identical: No visual changes
  • >99% similar: Minor differences (anti-aliasing, subpixel rendering) - usually a PASS
  • <99% similar: Meaningful changes detected - review the overlay

When similarity is lower than expected, view the overlay image to see exactly what changed. Correlate with your code changes to determine if the diff is intentional.

Options

--threshold, -t   Color difference threshold (0-1). Default: 0.1
                  Higher = more tolerant of minor color variations

--alpha, -a       Overlay tint transparency (0-1). Default: 0.5

--output, -o      Custom output path for overlay image

Fixes