Word Document Processing

Problem

A .docx file is a ZIP package of XML, so naive string replacement can miss w:ins, w:del, comments, and original RSID values, producing unreadable redlines or broken formatting. This skill targets engineers who need to create, read, and review Word documents in automated workflows, with a focus on verifiable output rather than ad hoc text substitution.

How It Works

• Read/analyze: extract Markdown with pandoc, using --track-changes=all to preserve tracked changes; when comments, media, or complex structure matter, unpack and inspect word/document.xml, word/comments.xml, and word/media/.
• Create documents: build and export .docx files with the Node docx library using Document, Paragraph, and TextRun components.
• Edit existing documents: plan the redline first, group changes into 3 to 10 batches by section or type, unpack the document, apply replacements with an OOXML document library, preserve original RSID for unchanged text, split replacements into unchanged + deleted + inserted + unchanged, repackage, and verify by converting back to Markdown.

Boundaries

It fits new documents, text extraction, legal/contract redlines, and structural analysis; complex layout tweaks are best confirmed in a desktop client. Output includes a DocArtifact and redline notes, suitable for downstream PDF conversion or multimodal understanding.