/screenshot

Take a screenshot that Claude can use for visual analysis. The image is automatically compressed for efficient context usage.

Usage

/screenshot [target] [--size SIZE] [--output PATH]

Arguments

Examples

# Interactive window capture (click on window)
/screenshot

# Capture specific application window
/screenshot Safari
/screenshot "Google Chrome"

# Capture region (drag to select)
/screenshot region

# Full screen capture
/screenshot screen

# Custom size limit (500KB for more detail)
/screenshot --size 500

# Custom output path
/screenshot --output ./debug/current-state.png

# Combined options
/screenshot Safari --size 100 --output ./browser.png

What Happens

1. The screenshot script runs based on your target selection
2. For window or no target: cursor becomes camera - click on desired window
3. For region: cursor becomes crosshair - drag to select area
4. For screen: captures entire main display
5. For app name: activates that app and captures its window
6. Image is auto-compressed if larger than size limit
7. File path is returned so Claude can read the image

After Capture

Claude will automatically read the captured image and can:

• Analyze UI layouts and designs
• Identify visual bugs or issues
• Describe what's shown on screen
• Compare against expected appearance

Notes

• macOS only - uses built-in screencapture
• Interactive captures (window/region) require user action
• Screenshots saved to ./screenshots/ by default
• Compression uses sips (built-in) or pngquant if available