Agent Skills: DOCX Processing

DOCX Processing

Workflow Decision Tree

• Reading/Analyzing: Use text extraction or raw XML access
• Creating New Document: Use docx-js (JavaScript)
• Editing Existing: Use OOXML editing or redlining workflow

Reading Content

Text Extraction with Pandoc

# Convert to markdown with tracked changes
pandoc --track-changes=all file.docx -o output.md

Raw XML Access

# Unpack document
unzip document.docx -d unpacked/
# Key files:
# word/document.xml - Main content
# word/comments.xml - Comments
# word/media/ - Images

Creating New Documents (docx-js)

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

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

const buffer = await Packer.toBuffer(doc);
fs.writeFileSync('document.docx', buffer);

Editing Existing Documents

Simple Edits

1. Unpack: unzip doc.docx -d unpacked/
2. Edit word/document.xml
3. Repack: cd unpacked && zip -r ../edited.docx .

Tracked Changes (Redlining)

For professional documents, use tracked changes:

<!-- Deletion -->
<w:del w:author="Author" w:date="2025-01-01T00:00:00Z">
  <w:r><w:delText>old text</w:delText></w:r>
</w:del>

<!-- Insertion -->
<w:ins w:author="Author" w:date="2025-01-01T00:00:00Z">
  <w:r><w:t>new text</w:t></w:r>
</w:ins>

Converting to Images

# DOCX to PDF
soffice --headless --convert-to pdf document.docx

# PDF to images
pdftoppm -jpeg -r 150 document.pdf page

Best Practices

• Use Pandoc for text extraction
• Use docx-js for creating new documents
• For legal/business docs, always use tracked changes
• Preserve original RSIDs when editing