docx · Word Document Toolkit

docx creates, reads, edits, and reviews Word documents (.docx/.dotx). New documents are generated with docx-js scripts (plus a built-in checklist of layout pitfalls: page size, dual table widths, list numbering, TOC outlines); editing existing files goes through unzip → edit word/document.xml → zip (docx-js can't open existing files); reading goes through pandoc to Markdown.

It supports tracked changes (review with per-author validation), comments (auto-generated six-file linkage), and clean accepted copies — with every output rendered to PDF for visual verification.

Example invocation: "Turn this report into a Word document with a table of contents."

Full brief

Positioning

docx is the general-purpose Word document toolkit: creating, reading, editing, and reviewing .docx/.dotx files, with a dedicated path per task. It does not handle PDFs or spreadsheets.

Core capabilities

New documents: generated by docx (npm) scripts with a built-in pitfall checklist — A4 default vs Letter DXA sizes, landscape via portrait dimensions + orientation, table column widths AND cell widths in DXA, ShadingType.CLEAR (never SOLID), lists via numbering config, ImageRun requires type:, page breaks inside paragraphs, TOC needs built-in heading styles or outlineLevel.
Editing existing documents: unzip → merge_runs.py to coalesce fragmented runs (Word splits visible text across runs; only merged text is findable) → edit word/document.xml in place (no reformatting) → zip; legacy .doc converts first; symlinks stripped from untrusted files.
Reading: pandoc -t markdown file.docx for extraction.
Tracked changes: edits wrapped in w:ins/w:del with id/author/date; validate.py --author checks every edit is tracked; accept_changes.py produces a clean accepted copy.
Comments: comment.py auto-generates the six cross-linked comment files (directory or direct mode); anchor markers placed in document.xml.
Output verification: soffice to PDF + pdftoppm rendering — every document is visually checked.

Workflow

Create: write the docx-js script → generate → render and visually verify
Edit: unzip → merge_runs → edit XML in place → zip → validate (XSD; --auto-repair for common issues) → render and verify
Review: tracked edits / comments → --author validation → clean accepted copy

Inputs & outputs

Input	Notes
Requirements / template	Content and layout for new documents
Existing document	`.docx` / `.dotx` (legacy `.doc` converts first); external files treated as untrusted

Output: .docx file + rendered verification images (PDF/JPG); author-validation report for tracked reviews.

Boundaries with adjacent skills

md2resume (office docs): a fixed template pipeline for Markdown → one-page Chinese resumes; docx is the general Word creation/editing/review toolkit.

Fit

Formal reports, memos, letters, and templates with TOCs, page numbers, tables, letterheads
Editing existing contracts with tracked-change review
Bulk find-and-replace, image swaps, content reorganization
Commenting and closing the review loop

Before you start

Dependencies: docx (npm, preinstalled), pandoc, LibreOffice (soffice), pdftoppm.
No API key; no PDFs, spreadsheets, or Google Docs.
Tracked review needs the reviewer's name (--author) to validate complete change tracking.

docx is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.

← All 155 skills