DOCX Document Toolkit

Problem

Many Word tasks get stuck when text extraction is not enough: structure, comments, tracked changes, embedded media, and metadata can be lost if you only copy text or edit raw XML blindly. This skill targets document workflows that need reliable .docx creation, editing, and analysis, turning the process into concrete paths instead of one-off conversion commands.

How It Works and Limits

• Reading: use pandoc to export Markdown for body text and revisions; unpack the package when comments, complex formatting, embedded media, or metadata matter.
• Creating: build files with docx-js using Document, Paragraph, and TextRun, then export with Packer.toBuffer().
• Editing: use the Python Document library for OOXML operations: unpack, modify, save, and repack. For documents from others, or legal, academic, business, and government documents, prefer the tracked-change workflow and batch 3 to 10 related changes.
• Visual enhancement: for new documents or complex concepts, describe schematics in natural language to generate publication-quality figures, saved under figures/, for workflows, architecture, data flow, and other relationships.
• Analyzing: when visual inspection helps, convert to PDF and render pages with pdftoppm to check layout, pagination, and structure.

Note: plain-text extraction can stay in Markdown, but preserving formatting, comments, and tracked changes requires XML-level work. Do not use Markdown line numbers to locate XML, because they change after edits.