mediar-ai/packages/skills/skills/docx-toolkit

DOCX Toolkit

Comprehensive document creation, editing, and analysis for Word documents (.docx) with support for tracked changes, comments, and formatting preservation.

Prerequisites

• pandoc (for text extraction)
• docx-js (npm) for creating new documents
• Python with defusedxml for editing

Instructions

Reading/Analyzing Content

1. Text Extraction with pandoc

   pandoc --track-changes=all document.docx -o output.md
   

2. Raw XML Access (for comments, formatting, metadata)

   python ooxml/scripts/unpack.py document.docx output_dir
   

   Key files: word/document.xml, word/comments.xml, word/media/

Creating New Documents

Use docx-js (JavaScript):

import { Document, Paragraph, TextRun } from 'docx';

const doc = new Document({
  sections: [{
    children: [
      new Paragraph({ children: [new TextRun("Hello World")] })
    ]
  }]
});

Editing Existing Documents

1. Unpack: python ooxml/scripts/unpack.py file.docx dir
2. Edit XML files using Document library
3. Pack: python ooxml/scripts/pack.py dir file.docx

Redlining Workflow (Tracked Changes)

For legal, academic, business docs - use minimal, precise edits:

# Only mark text that actually changes
# Preserve original run's RSID for unchanged text

1. Get markdown: pandoc --track-changes=all file.docx -o current.md
2. Identify and group changes (3-10 per batch)
3. Map text to XML, implement changes
4. Pack and verify

Error Handling

• If password-protected, use decryption tools first
• Validate XML after each edit before packing
• Test tracked changes in Word to verify rendering

Notes

• Tracked changes: <w:ins> (insertions), <w:del> (deletions)
• Convert to images: soffice → PDF → pdftoppm

Source: anthropics/skills