PDF Processing Skill (Open Cowork built-in)
A complete PDF playbook for AI agents: extract text and tables, create, merge, split, watermark, encrypt, and fill forms — so the AI stops guessing and starts using the right tool.
Also a bundled copy of Anthropic's official pdf skill (this site has separately written a detailed profile for the official anthropics/skills source) — content is trustworthy but the version may lag.
What does this skill do, and when should you use it?
This is the built-in pdf skill of the Open Cowork desktop app, located at .claude/skills/pdf/ in the repository. It ships as a SKILL.md guide rather than executable scripts: it teaches the agent to use Python libraries (pypdf, pdfplumber, reportlab) and CLI tools (qpdf, pdftotext, pdftk) for PDF work. Coverage includes text/table extraction, PDF creation, merge/split, rotation, metadata, OCR for scanned files, watermarking, and password encryption. Form filling and advanced topics (pdf-lib, pypdfium2) are deferred to separate FORMS.md and REFERENCE.md files. Note: while the repository is MIT-licensed, the SKILL.md itself is marked proprietary with terms in LICENSE.txt — verify before commercial adoption.
The skill guides the AI to: merge, split, extract metadata from, rotate, watermark, and password-encrypt PDFs with pypdf; extract layout-aware text and export tables to Excel with pdfplumber plus pandas; generate new PDFs via reportlab's Canvas and Platypus; OCR scanned documents using pytesseract and pdf2image; and perform extraction, merge, split, rotation, and decryption from the command line with pdftotext, pdfimages (poppler-utils), qpdf, and pdftk. Form-filling instructions live in a separate FORMS.md.
- An admin or finance assistant needs the AI to batch-extract tables from PDFs into Excel
- A developer wants multiple report PDFs merged or split page-by-page
- Programmatic generation of multi-page PDF reports with headings and body text
- A scanned PDF needs OCR to become searchable, extractable text
- Watermarking or bulk password-protecting outgoing PDFs
- Having the AI fill in a PDF form automatically (per FORMS.md)
What are this skill's strengths and limitations?
- Broad task coverage: extraction, creation, merge/split, OCR, encryption, watermarking, forms in one guide
- A task-to-tool recommendation table reduces agent tool-choice hesitation
- Copy-paste-ready code snippets lower the chance of agent errors
- Built on mature, mainstream tooling (pypdf, pdfplumber, reportlab, poppler)
- The skill is marked proprietary (LICENSE.txt), differing from the repo's MIT license — verify terms before commercial use
- No bundled scripts; it depends on multiple Python libraries and system tools being pre-installed
- OCR quality (Tesseract) is unspecified, particularly for non-English languages
- Open Cowork itself currently ships installers only for Windows and macOS; Linux requires building from source
- Form-filling and pdf-lib content lives in FORMS.md/REFERENCE.md, which are not provided here, so those details cannot be verified
How do you install this skill?
As a built-in Open Cowork skill, it ships with the app under .claude/skills/ — install the app and it is available. To use it in any other Agent Skills-compatible client, copy the .claude/skills/pdf/ folder (SKILL.md, FORMS.md, REFERENCE.md, LICENSE.txt) into your .claude/skills/ directory. The app itself installs via Homebrew (brew tap OpenCoworkAI/tap && brew install --cask --no-quarantine open-cowork), installers from the Releases page, or building from source. The Python libraries and CLI tools are prerequisites; the SKILL.md does not provide an automated setup step for them.
How do you use this skill?
Trigger with natural language, e.g. "Merge these three PDFs into one", "Extract all tables from this report and save as Excel", or "Fill out this PDF form". The AI picks the appropriate library or CLI tool per SKILL.md's task table. Ensure the required dependencies (pytesseract, qpdf, etc.) are installed in the runtime environment.
How does this skill compare with similar options?
Open Cowork positions itself as the open-source implementation of Claude Cowork, adding multi-model support, GUI automation, Feishu/Slack remote control, and VM-level sandbox isolation. Within Open Cowork, this pdf skill sits alongside sibling pptx, docx, and xlsx skills as the document-processing family.