name: docx
description: Use this skill whenever the user wants to create or edit .docx (Microsoft Word) files. This includes creating new Word documents from scratch, editing existing Word documents, extracting content from Word files, converting between formats, adding images or other content to Word files, or manipulating Word file structure. If the user mentions .docx, Word, DOCX, or wants to produce a Word document, use this skill.
license: Proprietary. LICENSE.txt has complete terms

DOCX Processing Guide

Overview

This guide covers creating, editing, and analyzing Word documents (.docx files). Use this skill when you need to programmatically create new Word documents, make targeted edits to existing files, or extract content.

Creating Documents

Use the docx JavaScript library for programmatic document creation. The library provides a powerful API for building Word documents with full formatting control.

Page Sizing

Always explicitly set page size — docx-js defaults to A4 (11,906 × 16,838 DXA), not US Letter. Specify dimensions using DXA units, where 1 inch = 1440 DXA.

For US Letter (8.5" × 11"):
- Width: 12,240 DXA
- Height: 15,840 DXA
- With standard 1" margins: content area is 10,240 × 13,840 DXA

For landscape orientation, pass portrait dimensions with landscape: true — the library handles internal conversion:

const doc = new Document({
  sections: [{
    properties: {
      pageHeight: 15840,  // Portrait height
      pageWidth: 12240,   // Portrait width
      landscape: true,    // Library rotates internally
    }
  }]
});

Tables

Tables require dual width specifications. Always specify both columnWidths array AND individual cell width properties, with all values using WidthType.DXA (percentage widths break in Google Docs).

Table width must equal the sum of column widths:

const columnWidths = [3000, 4000, 3000];  // Total: 10000
const table = new Table({
  columnWidths: columnWidths,
  width: { size: 10000, type: WidthType.DXA },  // Match total
  rows: [
    new TableRow({
      cells: [
        new TableCell({
          children: [...],
          width: { size: 3000, type: WidthType.DXA }
        }),
        // ... more cells
      ]
    })
  ]
});

Use ShadingType.CLEAR for cell background colors to avoid rendering issues.

Bullets and Lists

Never insert bullet points using manual Unicode characters. Instead, use the numbering system with LevelFormat.BULLET:

new Paragraph({
  text: "Item 1",
  bullet: {
    level: 0
  }
})

Smart Typography

Maintain professional typography with smart quotes and apostrophes via XML entities in the underlying document:
- “ — Left double quote (")
- ” — Right double quote (")
- ’ — Apostrophe/Right single quote (')

These are handled automatically when using standard string methods, but when editing raw XML, use these entities explicitly.

Editing Existing Documents

Use the three-step workflow: unpack → edit → pack.

Step 1: Unpack

Extract the .docx archive into editable XML:

unzip -q input.docx -d unpacked/

Step 2: Edit

Edit files in unpacked/word/document.xml directly. Prefer the Edit tool for transparent string replacement rather than writing custom Python scripts.

When applying tracked changes (revision marks), copy the original run's formatting block (<w:rPr>) to the new run to maintain style consistency:

<!-- Original run with formatting -->
<w:r>
  <w:rPr>
    <w:b/>
    <w:color w:val="FF0000"/>
  </w:rPr>
  <w:t>Original text</w:t>
</w:r>

<!-- New run with revision mark — copy the <w:rPr> block -->
<w:r>
  <w:ins w:id="0" w:author="Editor" w:date="2024-01-15T10:00:00Z"/>
  <w:rPr>
    <w:b/>
    <w:color w:val="FF0000"/>
  </w:rPr>
  <w:t>Edited text</w:t>
</w:r>

When deleting entire paragraphs, mark both the paragraph content AND the paragraph element itself as deleted to prevent empty rows from persisting after change acceptance:

<!-- Mark paragraph deletion at both content and element level -->
<w:del w:id="0" w:author="Editor" w:date="2024-01-15T10:00:00Z">
  <w:p>
    <w:pPr>
      <w:pStyle w:val="Normal"/>
    </w:pPr>
    <w:r>
      <w:annotationStart w:id="0"/>
      <w:t>Deleted paragraph</w:t>
      <w:annotationEnd w:id="0"/>
    </w:r>
  </w:p>
</w:del>

Step 3: Pack

Repack the XML back into a .docx file with validation:

cd unpacked/
zip -r -q ../output.docx word/ _rels/ '[Content_Types].xml' customXml/ 2>/dev/null
cd ..

If validation errors occur, the library has auto-repair capabilities — repack and try again.

Reading Documents

Extract Text

Use pandoc to extract plain text while preserving document structure:

pandoc input.docx -t plain -o output.txt

Access Raw XML

For programmatic content analysis, access the document's underlying XML structure directly via the unpacking method above.

Important Constraints

Line breaks — Never use \n for line breaks within a paragraph. Instead, create separate Paragraph elements or use explicit line break elements.

Page breaks — PageBreak elements must nest within Paragraph containers; they cannot exist at document root level.

ImageRun — When adding images, always specify the explicit type parameter (e.g., type: "picture").

Heading styles — Heading elements require outlineLevel attributes for table-of-contents generation:

new Paragraph({
  text: "Section Title",
  style: "Heading1",
  outlineLevel: 0
})

Built-in style overrides — Override built-in Word styles using exact identifiers (e.g., "Heading1", "Normal"), not custom names.

See REFERENCE.md for advanced techniques, more examples, and JavaScript library details.