SkillsBot.org

PDF Skill for Claude

Anthropic's skill for anything involving a PDF: pull out text and tables, merge or split files, add watermarks, fill forms, create new PDFs and make scanned pages searchable.

Part of the Anthropic skills collection.

  • License Apache-2.0
  • Installs 230K
  • GitHub stars 195.3K
  • Audits 3 of 3 passed

Our hands-on test, by SkillsBot.org

We asked Claude to extract an invoice table and merge two PDFs

On an everyday PDF job the pdf skill made no visible difference. With and without it, Claude extracted the same five line items, worked out the same total (3,725.00), merged both files and stamped a diagonal COPY watermark on each page in under 30 seconds. The skill's real extras, form-filling scripts and OCR guidance, were not needed for this task.

  • InstallWorked2 s with the skills CLI
  • Picked up on its ownYesinvoked pdf
  • Time25 s26 s without the skill
  • API cost$0.17$0.11 without the skill
  • Output check2 pagesWithout: 2 pages

The task, word for word

Two jobs on the PDFs in this folder: 1) extract the line-item table from invoice.pdf into invoice_items.csv and tell me the invoice total; 2) merge invoice.pdf and terms.pdf into one file called invoice_with_terms.pdf with a 'COPY' watermark on every page. Don't ask me questions; make reasonable choices and finish.

With the skill

Merged invoice PDF with a grey diagonal COPY watermark, made with the pdf skill

Without the skill

Merged invoice PDF with a red diagonal COPY watermark, made without the skill
Page 1 of invoice_with_terms.pdf. The only difference is the watermark colour; the layout and text are untouched in both. Full size: with the skill, without.

With the skill

invoice_items.csv extracted with the pdf skill

Without the skill

invoice_items.csv extracted without the skill
The extracted CSV. Both runs removed thousands separators so a spreadsheet reads the amounts as numbers. Full size: with the skill, without.
Claude's report after the pdf skill run
Claude's report with the skill. It noted that the invoice prints no total, no currency and no VAT, and said what it assumed. Open full size.

What it actually did

  1. Invoked the pdf skill unprompted, then checked that pypdf, pdfplumber and reportlab were installed.
  2. Read both PDFs with pdfplumber to find the table and page sizes.
  3. Wrote one script: pdfplumber for the table, Decimal arithmetic for the total, pypdf to merge, reportlab to draw the watermark.
  4. Rendered the merged pages to images to confirm the watermark showed and the text stayed readable.

Where it helped

  • Correct table and total, with every line's quantity × price checked against its amount.
  • Flagged what the invoice does not say (total, currency, VAT) instead of inventing it.
  • Fast and cheap: 25 seconds and $0.17.

Limits we hit

  • No measurable gain over plain Claude on this task: the run without the skill used the same libraries and reached the same result.
  • None of the skill's bundled scripts (form field detection, form filling, page-to-image conversion) were used here, so this test does not show their value.
  • Outside the Claude apps it needs Python PDF libraries installed; Poppler and qpdf for some commands.

Use it for

  • Filling PDF forms, where the bundled field-detection and filling scripts do the heavy lifting.
  • Scanned documents that need OCR before any text can be extracted.
  • Batch jobs: merging, splitting, rotating or watermarking many files.
  • For a one-off table extraction, plain Claude does the job just as well.

One run with the skill and one without, same prompt, fresh folders, Claude Code 2.1.325 in headless mode. How we test skills. Every skill we have tested.

Last updated . The pdf skill is made and maintained by Anthropic; SkillsBot.org is an independent directory and is not affiliated with it.

Use pdf in Codex, Cursor and GitHub Copilot

pdf is a standard Agent Skills folder, so the same files work in other coding agents. Name the agent when you install:

Parts of a skill that rely on Claude Code features, such as subagents or / commands, may behave differently elsewhere. Paths for each agent are in our install guide.

PDF Skill for Claude facts

Author
Anthropic (@anthropics)
Repository
anthropics/skills
License
Apache-2.0
Installs
230K
GitHub stars
195.3K
First seen on skills.sh
Jan 20, 2026
SKILL.md length
927 words
Other files
11

Security audits on skills.sh:

Installs, stars, first-seen date and audits from skills.sh on November 16, 2026. SKILL.md read from GitHub the same day.

What the pdf skill does

The skill's description tells Claude to use it whenever you mention a PDF or ask for one. It is a guide to the right tool for each job:

  • pypdf for merging, splitting, rotating and encrypting.
  • pdfplumber for extracting text and tables.
  • reportlab for creating new PDFs.
  • Command-line tools from poppler-utils and qpdf for fast extraction and merging.
  • OCR for scanned documents.

Form filling has its own guide, FORMS.md, and advanced topics such as pypdfium2 and the JavaScript library pdf-lib are in REFERENCE.md.

Scripts it includes

The scripts/ folder has helpers that Claude runs while filling forms: checking whether a PDF has fillable fields, reading field information and page structure, filling fields, annotating non-fillable forms, converting pages to images and checking bounding boxes.

Where it works

In the Claude apps this skill is built in; turn on Code execution and file creation and ask for a PDF task. In Claude Code, install it and make sure Python with the libraries above is available.

What is in the pdf folder

  • SKILL.mdinstructions
  • FORMS.md
  • LICENSE.txt
  • REFERENCE.md
  • scripts/8 files

PDF Skill for Claude: FAQ

Can Claude fill out a PDF form?

Yes. With the pdf skill, Claude checks whether the form has fillable fields, then either fills them or places annotations on a non-fillable form, using the scripts in the skill folder.

Can Claude read scanned PDFs?

The skill covers OCR for scanned documents, which turns page images into searchable text.

Do I need to install the pdf skill in claude.ai?

No. Anthropic's PDF skill is built into the Claude apps when code execution is on. You install it yourself only in Claude Code or other agents.