Documents

Use pdfplumber to detect and extract tables from each PDF page, then load the results into pandas DataFrames. Combine the tables and export them with df.to_excel() for a clean spreadsheet output.

Build the workbook with openpyxl, writing Excel formulas like =SUM() instead of hardcoded Python-calculated values. Apply color coding (blue inputs, black formulas, green cross-sheet links), then run recalc.py with LibreOffice to calculate values and verify zero formula errors.

Use pdf-lib in JavaScript or pypdf in Python to populate form fields, following the forms.md guide in the PDF sub-skill. The workflow reads existing form fields, fills them from your data, and saves a flattened completed PDF.

Yes, the DOCX sub-skill supports a redlining workflow using raw OOXML editing. You unpack the .docx, insert w:ins and w:del revision elements in the document XML, then repack so changes appear as tracked revisions in Word.

These errors come from invalid cell references or division by zero in your formulas. Run recalc.py to get a JSON report listing each error type and location, fix the referenced cells or guard denominators, then recalculate again until zero errors remain.

Use Gemini 3 Pro for large, complex PDFs mixing text, images, charts, and diagrams, since it processes the file multimodally within a 1M token context. Use pdfplumber for simpler text-and-table extraction where visual understanding is not needed.