Extract PDF

Extract structured text from PDF files while preserving heading hierarchy, list structure, and table layout where possible. Wraps a local pdf-parsing library — never sends bytes off-device.

When to use this skill

Reach for extract-pdf when the user asks you to read, summarize, or extract content from a PDF file. Don't use this skill for structured data extraction — pair with extract-table downstream for that.

Inputs

• A path to a .pdf file under the user's working directory
• Optional: --format=json to emit structured output
• Optional: --pages=1-5 to scope extraction to a page range

Outputs

• Plain text by default. Headings preserved as # markers, lists as - bullets
• JSON when --format=json — includes pages, headings, paragraphs, and tables

Examples

Basic extraction

User: "What's in spec-v2.pdf?" → run extract-pdf ./spec-v2.pdf, then summarize the result.

Page range

User: "Read the first chapter of book.pdf" → estimate page range from the table of contents, run extract-pdf ./book.pdf --pages=1-22.

Caveats

• Scanned PDFs without an embedded text layer return empty. Pair with ocr-image for image-based PDFs.
• Tables in irregular layouts may extract as flat text.
• Encrypted PDFs require the password as a third positional argument.