How to Install and Use the PDF Skill with Claude Code (2026)

PDFs are where useful information goes to get stuck. A contract you need one clause from. A bank statement with a table you want in a spreadsheet. A scanned form someone emailed you that you can't even select text in. Copy-paste either mangles the layout or doesn't work at all.

The PDF skill for Claude Code handles all of that from the terminal. It reads text and tables, merges and splits files, fills forms, adds watermarks, pulls out images, and runs OCR on scans so they become searchable. Installing it takes about two minutes. This walks through the whole thing, plus the part most guides skip: what to actually ask it once it's installed.

What the PDF skill does

It's worth knowing the range before you install, because it's wider than "read a PDF." The skill covers:

• Extracting text and tables, keeping the table structure intact
• Merging several PDFs into one, or splitting one into parts
• Rotating pages and reordering them
• Adding watermarks
• Filling out PDF forms
• Encrypting and decrypting files
• Extracting embedded images
• Running OCR on scanned, image-only PDFs to make them searchable
• Creating new PDFs from scratch

If your task is on that list, the skill beats pasting text into a chat window, because it works on the file directly and handles the cases where copy-paste falls apart.

Before you start

You need two things: Claude Code installed and working, and a terminal you're comfortable running a couple of commands in. That's the whole prerequisite list. If you're already using Claude Code, you're ready.

Step 1: Find your skills directory

Claude Code loads skills from a folder in your config directory. On Mac and Linux that's ~/.claude/skills/. On Windows it's %USERPROFILE%\.claude\skills\.

# Mac / Linux
cd ~/.claude/skills/

# If it doesn't exist yet
mkdir -p ~/.claude/skills/ && cd ~/.claude/skills/

Step 2: Install the skill

A skill is just a folder with a SKILL.md inside it. Installing means getting that folder into your skills directory. The PDF skill ships in Anthropic's skills collection, so you clone the collection and copy the one folder out:

# From inside ~/.claude/skills/
git clone https://github.com/anthropics/skills.git
cp -r skills/pdf ~/.claude/skills/pdf

If you got the skill another way, a download or a marketplace, the goal is identical: end up with a folder containing SKILL.md inside ~/.claude/skills/. The folder name and that file are all Claude needs.

Some PDF operations rely on helper libraries the skill calls under the hood. If Claude reports a missing dependency the first time you run a task, it will usually tell you the exact install command, let it install what it needs, or run the command yourself.

Step 3: Verify it loaded

Restart Claude Code, or start a fresh session, so it re-scans the skills folder. Then ask:

Which skills do you have available? Is the PDF skill one of them?

Claude reads each skill's description on startup and loads the full instructions only when a task matches. If it lists the PDF skill, you're done installing, it'll fire automatically the next time you ask something PDF-shaped.

Step 4: Run your first task

Here's where the skill earns its place. Some prompts that work well:

Pull a table out of a statement into a spreadsheet:

Extract the transactions table from statement-may.pdf and save it as a CSV,
with columns for date, description, and amount.

Merge a stack of files:

Merge invoice-01.pdf through invoice-12.pdf into one file called 2026-invoices.pdf,
in order.

Make a scan searchable:

Run OCR on scanned-contract.pdf and give me the text of clause 7.

Notice the pattern: name the file, say exactly what you want out, and say the format. A vague "read this PDF" gets a vague answer. A specific request gets you the table, the clause, or the merged file.

Step 5: Check the output on anything that matters

The skill is good, not infallible, and two cases need a human eye. OCR on a poor scan can misread a digit, so verify numbers you're going to act on. And table extraction from a weird layout can occasionally merge or split a column, so glance at the structure before you trust it downstream. For a clean, born-digital PDF, extraction is reliable. For a faxed-then-scanned-then-emailed mess, read it back.

Three jobs it quietly nails

Beyond reading and merging, a few tasks are worth knowing about because they save the most tedious work.

Filling forms. If you get the same PDF form repeatedly, a reimbursement sheet, an intake form, you can have the skill fill the fields from data you provide instead of clicking through the form by hand each time. Tell it the field values and the file, and it writes them in.

Splitting a big document by section. A 200-page report you only need three chapters from doesn't need to travel as 200 pages. Ask the skill to split it at the page ranges you name, or by its bookmarks if it has them, and you get clean, smaller files. The same works in reverse for stapling a stack of one-pagers into a single document in a set order.

Stripping a file down for sharing. Watermarking a draft before it goes out, removing pages you don't want a client to see, flattening a form so the values can't be edited, these are the small finishing tasks that usually mean opening a paid PDF editor. The skill does them from the terminal in one prompt.

None of these is glamorous. All of them are the kind of five-minute task you do often enough that automating it adds up.

When the PDF skill struggles

It's worth knowing the limits before you rely on it. Heavily designed PDFs, magazine layouts, multi-column academic papers, infographics, can confuse text extraction because the reading order isn't obvious to a machine. You'll usually still get the text, just not always in the order a human would read it. Scanned documents with handwriting are hit or miss; printed text OCRs well, cursive does not. And very large files (hundreds of megabytes of images) can be slow. For the common case, a born-digital PDF with normal text and tables, none of this bites. Knowing where it gets shaky just tells you when to read the output twice.

Chain it with the Word and Excel skills

The PDF skill is at its best as one link in a chain. Extract a table from a PDF, then hand it to the spreadsheet skill to build a formatted .xlsx with formulas. Or pull the body of a report out of a PDF and have the Word skill lay it into a clean .docx with headings and a table of contents. Because each skill is a separate folder, you install the ones you need and they cooperate when a task spans formats.

The security note you shouldn't skip

A skill is instructions, and sometimes code, that run with your permissions. Before installing any skill, open its SKILL.md and read it. It's plain text, so you can see exactly what it tells Claude to do and which tools or services it touches. For a PDF skill, what you're checking is simple: it should work on your files locally and not ship them off to some external service it doesn't need. Install from the official Anthropic repo when you can, and audit anything from a source you don't recognize. We go deeper in our guide to auditing skill security.

Worth it?

If you touch PDFs more than once a week, yes. The first time it pulls a clean table out of a statement that would've taken you twenty minutes to retype, or makes a scanned document searchable, it pays for the two-minute install. And like every skill, it sits quiet until you ask for it, so there's no cost to having it ready. New to skills entirely? Start with how to install your first Claude skill, then come back for this one.

FAQ

What can the PDF skill actually do?

It reads and extracts text and tables from PDFs, merges and splits files, rotates pages, adds watermarks, fills forms, extracts images, and runs OCR on scanned documents so they become searchable. It can also create new PDFs.

Does the PDF skill work on scanned documents?

Yes. For scanned or image-only PDFs it runs OCR to pull the text out. Quality depends on the scan, so check the result on anything important.

Is my PDF sent anywhere when I use the skill?

The processing runs locally through the tools the skill calls. Read the SKILL.md before installing to confirm it does not upload your files to an external service. For document work it should not need to.

How is the PDF skill different from just pasting text into Claude?

Pasting works for short, clean text. The skill handles the messy cases, large files, tables that lose structure on copy-paste, scanned pages, and tasks like merging or form-filling that you cannot do by pasting.

Can I use the PDF skill and the Word or Excel skills together?

Yes. A common chain is extracting a table from a PDF and writing it into an .xlsx with the spreadsheet skill, or pulling PDF content into a formatted .docx.