pdf

---
name: pdf
description: Use when a task involves a PDF file — extracting text or tables, merging or splitting, rotating, watermarking, filling forms, encrypting, OCR, or creating one — so the agent picks the right library or CLI tool.
source: https://github.com/anthropics/skills/tree/main/skills/pdf
generated: 2026-07-02T18:42:59.795Z
category: tool
audience: general
---

## When to use

- Extracting text or tables from a PDF, including scanned pages that need OCR
- Merging, splitting, rotating, or watermarking pages across PDF files
- Creating a new PDF from scratch or filling out a PDF form
- Encrypting, decrypting, or extracting embedded images from a PDF

## Key concepts

### pypdf for structural edits

The go-to Python library for merging, splitting, rotating, watermarking, and encrypting — you copy pages between a PdfReader and PdfWriter and write the result out.

### pdfplumber for extraction

Used to pull text with layout and to extract tables (page.extract_tables), which can be loaded into pandas DataFrames and exported to Excel.

### reportlab for creation

Builds new PDFs either with a low-level Canvas or the higher-level Platypus flow (SimpleDocTemplate, Paragraph, Spacer, PageBreak) for multi-page documents.

### OCR for scanned PDFs

Scanned documents are converted to images with pdf2image, then run through pytesseract to make the text searchable/extractable.

### Command-line tools

poppler-utils (pdftotext, pdfimages), qpdf, and pdftk handle merge/split/rotate/decrypt from the shell when a Python round-trip isn't warranted.

### Progressive references

SKILL.md is the quick start; REFERENCE.md covers advanced pypdfium2 and JavaScript (pdf-lib) usage, and FORMS.md must be read before filling any PDF form.

## API reference

```
npx skills add anthropics/skills --skill pdf
```

Install the PDF processing skill.

```
npx skills add anthropics/skills --skill pdf
```

```
PdfReader / PdfWriter (pypdf)
```

Read pages and assemble output for merge, split, rotate, watermark, and encrypt operations.

```
from pypdf import PdfReader, PdfWriter
writer = PdfWriter()
for f in ["doc1.pdf", "doc2.pdf"]:
    for page in PdfReader(f).pages:
        writer.add_page(page)
with open("merged.pdf", "wb") as out:
    writer.write(out)
```

```
qpdf --empty --pages ... -- out.pdf
```

Merge, split, rotate, or decrypt PDFs from the command line.

```
qpdf --empty --pages file1.pdf file2.pdf -- merged.pdf
qpdf --password=pw --decrypt encrypted.pdf decrypted.pdf
```

## Gotchas

- Never use Unicode subscript/superscript glyphs in reportlab — the built-in fonts lack them and they render as black boxes; use <sub>/<super> tags in Paragraphs instead
- Filling a PDF form requires reading FORMS.md first; don't wing it with generic page edits
- Scanned PDFs have no extractable text layer — you must OCR via pdf2image + pytesseract before extraction works
- Table extraction quality depends on pdfplumber over pypdf; pypdf's extract_text loses tabular structure
- OCR and image conversion need system deps (tesseract, poppler) beyond the pip packages

---
Generated by SkillMake from https://github.com/anthropics/skills/tree/main/skills/pdf on 2026-07-02T18:42:59.795Z.
Verify against source before relying on details.

File: ~/.claude/skills/pdf/SKILL.md