How to use the PDF Processing Guide agent skill

What does the pdf skill make the agent do?

The skill covers the common PDF jobs: reading or extracting text and tables, combining or merging several PDFs, splitting documents apart, rotating pages, adding watermarks, creating new PDFs, extracting images, encrypting or decrypting files, and running OCR on scanned pages so they become searchable.

For PDF forms, the skill tells the agent to read a separate guide, forms.md, and uses helper scripts that check whether a form has fillable fields, convert pages to images, and place entries onto the form.

• Merge and split: combine files or cut a document into one file per page.
• Text and tables: pull text out of pages and tables into spreadsheet-style data.
• Pages and images: rotate pages, add watermarks, and extract embedded images.
• Scanned files: OCR pages so the text can be searched.
• Forms: check for fillable fields, then fill them using the bundled scripts.

How do you install the pdf skill?

Download the ZIP file from this page. You then have two options depending on how you use your agent.

For Claude Code, unzip the download into ~/.claude/skills/ to make the skill available everywhere, or into a project's .claude/skills/ folder to limit it to that project.

In the Claude apps, open Settings, go to Capabilities, and upload the skill there as a Skills file.

• Personal install: unzip into ~/.claude/skills/.
• Project install: unzip into the project's .claude/skills/ folder.
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP.

What do you need installed first?

The skill's instructions call Python libraries and command-line tools. Most steps fail until these are set up, so install them before asking for PDF work. The skill lists these requirements: Python, pypdf, pdfplumber, reportlab, pdf2image, pytesseract with the Tesseract OCR engine, pandas, Pillow, poppler-utils (which provides pdftotext and pdfimages), qpdf, and pdftk as an optional extra.

The skill's own text shows the pip install for pytesseract and pdf2image as a comment in its OCR example, so you can install those with pip. Tesseract and poppler are separate programs that you install with your operating system's tools.

• Python with pypdf, pdfplumber, reportlab, pdf2image, pytesseract, pandas and Pillow.
• Tesseract OCR engine, needed for scanned-page OCR.
• poppler-utils for pdftotext and pdfimages.
• qpdf for merging, splitting, rotating and decrypting from the command line.
• pdftk is optional.

How do you trigger the skill with a natural request?

You do not need a special command. The skill is meant to activate when you ask for something PDF-related, especially when your message mentions a .pdf file or asks for one to be produced. Describe the goal in plain language and name the file.

Be specific about the outcome. Say which files, which pages, and what you want back, such as a merged file or a spreadsheet of tables.

• "Merge invoice1.pdf, invoice2.pdf and invoice3.pdf into one file called q3-invoices.pdf."
• "Split report.pdf so each page is its own file."
• "Pull the tables from annual-report.pdf into an Excel file."
• "Rotate page 1 of scan.pdf by 90 degrees."
• "Make scanned-contract.pdf searchable and give me the text."

What are some practical examples?

Quick text extraction from a whole document uses pdftotext with a page range, which is useful for long reports where you only need a few pages.

Splitting and merging with qpdf works from the command line and is handy for batch jobs. The skill shows merging several files into one and splitting a file into page ranges.

Extracting images uses pdfimages, which writes images with a prefix you choose, such as output_prefix-000.jpg.

What about safety, scripts and privacy?

The skill runs scripts on your machine, and the form helpers are scripts you can read. The skill itself does not send data over the internet. Your files are processed by whatever tools the agent runs, so if the agent is working in a cloud environment, the content of your PDF goes through that environment. Do not use it on sensitive or confidential documents unless you are comfortable with that.

The scripts are example code. Their results are not verified, so do not rely on them for certified or legally binding form handling. Check important outputs yourself.

• Scripts run locally as the agent executes them; review them first if you are unsure.
• Check the results of form filling and OCR before you use them.
• The skill's license is listed as Proprietary, and LICENSE.txt holds the full terms; read it before redistributing the files.

What tips help the most?

Install the requirements first, then test with a copy of your file. For forms, ask the agent to follow forms.md in order, because the guide says to complete its steps in sequence before writing code. If the agent needs more depth, the skill points to reference.md for advanced libraries such as pypdfium2.

• Test on a copy of the PDF before running batch operations.
• For forms, let the agent run the fillable-field check first.
• If a scanned page gives poor text, check that Tesseract is installed and that the page image is clear.
• Use reference.md for advanced rendering and extraction options.