PDF Document Skill

Anthropic's official PDF skill teaches Claude to do almost anything with PDF files — extracting text and tables, merging, splitting, rotating, watermarking, creating new PDFs, filling forms, encrypting, extracting images, and running OCR on scans.

What is the PDF skill?

The PDF skill is one of the document skills that powers Claude's file capabilities. It packages the right Python libraries and command-line tools for each PDF task, plus the workflows to use them correctly. The core toolkit is pypdf for page operations, pdfplumber for text and table extraction, reportlab for creating PDFs, qpdf and pdftk for command-line manipulation, and pytesseract with pdf2image for OCR. It also references FORMS.md for filling PDF forms and REFERENCE.md for advanced libraries like pdf-lib and pypdfium2.

What it does

• Extracts text and tables from PDFs (pdfplumber's extract_text() and extract_tables()), including export to Excel
• Merges multiple PDFs into one and splits a PDF into individual pages
• Rotates pages, adds watermarks, and extracts images with pdfimages
• Creates new PDFs with reportlab, including multi-page reports and proper subscripts/superscripts
• Fills PDF forms and encrypts/decrypts files with user and owner passwords
• Runs OCR on scanned PDFs with pytesseract to make them searchable

How it works

Claude loads the skill's SKILL.md, which maps each task to the best tool. For reading, it opens the file with pypdf or pdfplumber and iterates pages; for tables it converts extracted rows into pandas DataFrames. For merging and splitting it uses pypdf's PdfWriter, or qpdf --empty --pages on the command line. Watermarks are applied by merging a watermark page onto every page. Creation uses reportlab's canvas or the Platypus document model, and the skill warns against unicode subscript/superscript glyphs (which render as black boxes) in favor of <sub> and <super> markup. Scanned documents are converted to images with pdf2image and read with pytesseract. Encryption uses pypdf's writer.encrypt(); decryption uses qpdf with the password.

Who it's for

Anyone who uses Claude to work with PDFs: analysts pulling tables out of financial statements, ops teams merging and splitting document bundles, legal teams encrypting or redacting files, and people who need to OCR scanned contracts or receipts into searchable text. It runs wherever Claude has a code/file environment — Claude.ai (paid plans), Claude Code, and the Claude API.

What you can build

• Automated pipelines that extract tables from PDFs into Excel
• Document assembly that merges, splits, and rotates pages on demand
• Watermarked or password-encrypted versions of sensitive files
• Searchable archives by OCR-ing scanned PDFs
• Generated reports and filled PDF forms from structured data

Why it matters

PDFs are the universal exchange format, but they are notoriously hard to parse and edit reliably — scanned pages have no text layer, tables lose their structure, and forms and encryption add friction. This skill encodes the tool-for-the-job choices Anthropic uses in production, so Claude reaches for pdfplumber for tables, pytesseract for scans, and qpdf for encryption instead of guessing. It turns messy PDF work into dependable, repeatable operations.

1. Install in Claude Code

Register the Anthropic skills marketplace, then install the document skills plugin:

/plugin marketplace add anthropics/skills
/plugin install document-skills@anthropic-agent-skills

Or install just this skill with the skills CLI:

npx skills add anthropics/skills --skill pdf

2. Use it

On Claude.ai paid plans these document skills are already available. Just ask:

Use the PDF skill to extract the tables from report.pdf into Excel.

Or: "OCR scanned-contract.pdf so the text is searchable, then merge it with appendix.pdf."