PDF Extract

Extract text/structure with provenance (pdf-extract)

Extract text/structure with provenance

Part of the Universal PDF Management & Creation Skill Repository. Category: PDF Management. Every skill is individually callable (INDIVIDUAL MODE) and may also be invoked automatically by the orchestrator (SYSTEM MODE).

When to use

Extract structured content while keeping page & position so downstream steps can cite sources.

Operations

• extract text
• per-page blocks with bbox
• multi-column via layout
• preserve page/position info
• image-only pages routed to OCR

Engine (what actually executes this)

universal_pdf/core.extract:extract_text / extract_blocks_per_page / extract_per_page

All heavy lifting is done on disk / one page at a time by the shared engine — never by loading the document into an LLM context.

Tools & graceful degradation

• Preferred tool: PyMuPDF text (text/dict)
• Fallback: pdfplumber (column layout)

If the preferred backend is unavailable the skill transparently uses the fallback or reports a documented limitation. It never reports success when an operation did not actually happen.

Validation (this skill attaches checks to its Outcome)

text_coverage; page coverage

Related skills

pdf-read, pdf-tables, pdf-ocr, pdf-provenance

Example requests it handles

• "Extract all the body text of this report keeping page numbers."

Notes

Provenance: outputs cite their source page/section where relevant.