PaddleOCR's document parsing goes beyond basic OCR by returning structured data including tables, mathematical formulas, charts, and full layout analysis. This is what you reach for when dealing with complex documents like financial reports, academic papers, or multi-column layouts where you need to preserve the document structure, not just extract raw text. The skill is strict about routing everything through the PaddleOCR API and explicitly forbids fallback attempts, which is honest but means you need the API configured before it's useful. Output comes as JSON with complete content in markdown format, and the docs are clear about displaying full results rather than truncating.
npx skills add https://github.com/freestylefly/canghe-skills --skill paddleocr-doc-parsing