PDF Screenshot

PDF Screenshot

This skill renders pages or specific regions (bounding boxes) of a PDF into PNG images. It is designed for human-agent collaboration workflows where the agent needs to show the human what it has detected (e.g., tables, figures) for verification or calibration.

Usage

Full Page Screenshot

/pdf-screenshot /path/to/doc.pdf --page 5

Saves to /tmp/pdf-screenshot/doc_page5.png (or custom path).

Cropped Region (Bounding Box)

/pdf-screenshot /path/to/doc.pdf --page 5 --bbox "72,200,540,400"

Saves to /tmp/pdf-screenshot/doc_page5_crop.png. Bounding box format: x0, y0, x1, y1. A default padding of 20px is added to the crop.

Custom Output Path

/pdf-screenshot /path/to/doc.pdf --page 5 --out /tmp/my_screenshot.png

Highlight Region

/pdf-screenshot doc.pdf --page 5 --highlight "72,200,540,400"

Draws a red rectangle around the specified region. Useful for visual verification without cropping.

Batch Processing

# Specific pages
/pdf-screenshot doc.pdf --pages 1,3,5 --out ./screenshots/

# All pages
/pdf-screenshot doc.pdf --all --out ./screenshots/

If --out is a directory (ends in /), files are named automatically (e.g., doc_page1.png).

Arguments

• pdf_path: Path to the input PDF file.
• --page: Single page number (0-indexed).
• --pages: Comma-separated list of page numbers (e.g. 1,3,5).
• --all: Process all pages in the document.
• --bbox: Optional crop bounding box x0,y0,x1,y1.
• --highlight: Optional highlight bounding box x0,y0,x1,y1 (red stroke).
• --out: Optional output path or directory.
• --dpi: Rendering DPI (default 150).

Dependencies

• pymupdf (fitz)