DOCX Toolkit — Create, Edit, and Analyze Word Documents

The problem

Programmatically generating and editing Word documents is notoriously fragile, often resulting in corrupted XML or inconsistent formatting across different viewers. Manually managing docx-js configurations or parsing raw OpenXML is time-consuming and prone to schema violations.

What it does

• Generates professional .docx files using docx-js with explicit control over page sizes, margins, and multi-column layouts.
• Edits existing documents by unpacking XML, merging adjacent runs, and repacking with automatic schema validation and repair.
• Implements complex document features including tracked changes, nested comments, table of contents, and dot-leader tab stops.
• Handles legacy .doc to .docx conversion and high-fidelity text extraction via Pandoc integration.
• Converts documents to PDF or high-resolution images using headless LibreOffice and Poppler.

Frameworks & tools

Node.js (docx-js), Python (scripts/office), Pandoc, LibreOffice, and Poppler.

Why this beats prompting it yourself

This skill enforces strict XML compliance rules that LLMs typically hallucinate, such as element ordering in pPr and proper namespace handling for tracked changes. It provides a structured workflow for unpacking and repacking files that prevents the common "corrupted file" errors associated with direct string manipulation of OpenXML.

Use cases

• Automate the generation of legal contracts with precise formatting and smart quotes.
• Review documents programmatically by inserting tracked changes and replies to comments as a named author.
• Build reporting pipelines that convert dynamic data into branded Word templates with headers, footers, and tables.
• Extract clean Markdown from complex Word documents while preserving tracked changes for analysis.

Known limitations

Requires a Node.js environment for new document creation and Python for the unpacking/packing utility scripts. Does not support direct manipulation of Google Docs or PDFs without local conversion.