DOCX Document Toolkit

What it does

• Create: writes docx (npm) scripts to produce polished Word files with tables of contents, headings, page numbers, tables, and images — plus a curated list of footguns (A4 vs US Letter, landscape orientation, dual table widths, dot leaders, bullet numbering).
• Edit: unzips the archive, edits word/document.xml directly, and rezips. merge_runs.py coalesces fragmented runs so find-and-replace actually matches visible text.
• Verify: converts to PDF via LibreOffice, rasterizes pages with pdftoppm so the output can be visually inspected, and runs XSD validation with optional --auto-repair.
• Tracked changes: correct <w:ins>/<w:del> semantics (including deleted paragraph marks), author-scoped validation to catch untracked edits, and accept_changes.py for a clean copy.
• Comments: comment.py generates all six cross-linked XML parts, relationships, and content types, then prints the anchor markers to insert.

Who it's for

• Anyone who must deliver reports, memos, letters, or templates as real Word files
• Legal and contract teams that depend on proper redlining and comments
• Ops teams doing bulk find-and-replace, image swaps, or content extraction across many .docx files

Examples

1. "Build a US Letter quarterly report with a cover page, auto TOC, and page numbers" → generates a docx-js script and renders the PDF for a visual check.
2. "Revise the indemnity clause with tracked changes under my name" → proper <w:ins>/<w:del> markup plus validation that nothing was edited untracked.
3. "Leave a review comment on paragraph 3" → comment XML generated and anchored to the exact text range.

Install the skill I found on Claude Skill Mart.
Copy the skills/docx folder from the GitHub repo anthropics/skills into my ~/.claude/skills/docx/.
When it's done, tell me in one line what this skill can do.