Anthropic PDF Skill

The pdf skill covers every PDF operation you’d otherwise script with half a dozen Python libraries. It picks the right tool — pypdf, PyMuPDF, or shell — based on what you’re trying to do.

What it can do

• Read & extract — text, tables, metadata, images
• Combine & split — merge multiple PDFs, or split one into per-page files
• Annotate — watermarks, page numbers, headers
• Forms — fill PDF forms programmatically (see FORMS.md)
• Security — encrypt, decrypt, password protect
• OCR — make scanned PDFs searchable
• Generate — create new PDFs from scratch with layout control

What’s surprising

• It defaults to pypdf for fast operations and only escalates to PyMuPDF when you need rendering or precise positioning
• Forms support is real — including filling, flattening, and verifying
• OCR can be applied selectively (only scanned pages)

Composes with

• pptx — convert decks to handout PDFs
• docx — convert reports to PDF for distribution
• brand-guidelines — apply Anthropic styling to the final PDF

Notes from curation

If your task involves the string “.pdf” anywhere, start here. Most users underestimate how much of “knowledge work” is actually PDF wrangling — this skill is the one that makes that part disappear.

Use the PDF skill to merge these three Q4 reports into one,
add a "CONFIDENTIAL" watermark on every page, and OCR the scanned
appendix so it becomes searchable.