Productivity & Collaboration pdfdocumentsocrformsofficial

PDF Processing

Read, generate, merge, split, watermark, encrypt, and OCR PDF files directly through Claude

What it does & when to use it

This is Anthropic's own PDF-handling skill — the same implementation that powers Claude.ai's file capabilities. It packages a mature toolchain (pypdf, pdfplumber, reportlab, qpdf, pdftotext, pytesseract) into a playbook Claude can follow, covering the vast majority of everyday PDF operations.

Extracts text and tables from PDFs, including OCR on scanned documents; merges, splits, and rotates pages; generates new multi-page PDFs from scratch; adds watermarks, encrypts/decrypts, and password-protects existing PDFs; extracts embedded images; fills out PDF forms (with a dedicated FORMS.md for form-specific guidance). Both command-line tools (qpdf, pdftk, pdftotext, pdfimages) and Python libraries are covered, and Claude picks whichever fits the task.

  1. Merge several scanned contracts into one PDF and add a watermark
  2. OCR text and tables out of a scanned invoice (image-based PDF)
  3. Batch password-protect a folder of PDFs before distribution
  4. Generate a multi-page report PDF from scratch in Python, with headers/footers
  5. Auto-fill a PDF application form

Pros & cons

Pros
  • Broad coverage — read/generate/edit/encrypt/OCR/forms are all handled with concrete examples
  • Both CLI and Python code snippets are given directly, easy to copy and adapt
  • Officially maintained and is literally the same implementation behind Claude.ai's file features, so behavior is predictable
Limitations
  • Licensed as Proprietary (source-available, not Apache/MIT) — you can reference it, but can't freely redistribute or repackage it commercially the way you could a community skill
  • OCR relies on pytesseract + pdf2image; on low-quality scans, accuracy is limited by those tools, not something the skill itself fixes
  • Complex PDF forms with nested or non-standard fields may still need manual fixing

How to install

Three options, pick whichever fits your setup:

  1. Claude Code: run /plugin marketplace add anthropics/skills, then /plugin install document-skills@anthropic-agent-skills (this plugin bundles pdf, docx, pptx, and xlsx together)
  2. Claude.ai: already enabled by default on paid plans, no install needed
  3. Claude API: upload the official skill (or your own) via the Skills API — see Anthropic's Skills API Quickstart docs

How to use

No manual trigger needed — just describe what you want done with a PDF in plain language and Claude picks up the skill automatically. For example: "Merge these three PDFs into one and rotate the first page 90 degrees," or "Extract the text from this scanned contract."

Compared to similar skills

Compared to hand-writing pypdf/pdfplumber code yourself, the value here is that the "which library for which task" judgment calls are already made (e.g. use pdfplumber, not pypdf, for table extraction; use qpdf for batch CLI operations), so Claude wastes less time trial-and-erroring. It ships alongside docx/pptx/xlsx in the same repo and packaging — install all four together if you need cross-format document handling.

FAQ

Is this skill open source?
No. It's licensed as Proprietary — see LICENSE.txt in the repo for full terms. Anthropic shared it as a reference implementation, not for free redistribution.
Is this the same thing as Claude.ai's built-in "upload a PDF" feature?
Same underlying implementation — the repo's README states these document skills are literally what powers Claude's file capabilities.
Do I need to install a lot of dependencies locally?
Depends on the operation: pure-Python ones (pypdf/pdfplumber/reportlab) just need pip installs; CLI operations need poppler-utils (pdftotext/pdfimages) plus qpdf or pdftk; OCR additionally needs pytesseract and poppler. These are usually pre-configured in Claude Code / Claude's own sandboxed environments.

Related skills