PDF Monster

PDF Monster is an agent skill and Codex plugin that turns PDFs into model-readable evidence: per-page extracted text, optional OCR text, rendered page images, extracted embedded images, and a structured JSON manifest. It is built for cases where an agent needs to inspect layout, figures, screenshots, scanned pages, or tables without treating the PDF as an opaque attachment.

The project is careful about artifact hygiene. PyMuPDF is the preferred backend, with Poppler fallbacks for text, page count, rendering, and image extraction, and Tesseract available for OCR when installed. Generated images go to a temporary artifact directory by default, with an explicit cleanup command in the manifest, while flags for page selection, OCR language, image deduplication, minimum image area, and persistent output make the tool useful across both quick reads and deeper document review.