Agent: Word Accessibility

word-accessibility - Microsoft Word (DOCX) Accessibility

Scans Microsoft Word documents for accessibility issues. Uses the scan_office_document MCP tool to parse DOCX files (ZIP/XML structure) and check for tagged content, alt text on images, heading structure, table markup, reading order, language settings, and color-only formatting.

When to Use It

• Reviewing Word documents before publishing or distributing
• Checking templates for accessibility compliance
• Auditing existing DOCX files as part of a document accessibility program
• Preparing documents for PDF conversion (accessibility issues carry over)

What It Catches

Expand - 11 Word accessibility rules (DOCX-E001 - DOCX-W005)

Rule	Severity	Description
DOCX-E001	Error	Images without alt text
DOCX-E002	Error	Missing document title in properties
DOCX-E003	Error	No headings used for document structure
DOCX-E004	Error	Tables without header rows
DOCX-E005	Error	Missing document language
DOCX-E006	Error	Color-only formatting conveying meaning
DOCX-W001	Warning	Very long alt text that needs summarization
DOCX-W002	Warning	Skipped heading levels
DOCX-W003	Warning	Merged cells in tables
DOCX-W004	Warning	Small font sizes below 10pt
DOCX-W005	Warning	Empty paragraphs used for spacing

Example Prompts

Show example prompts

/word-accessibility scan report.docx for accessibility issues
@word-accessibility review the quarterly report template
@word-accessibility check all Word documents in the docs/ directory

How to Launch It

In Claude Code (terminal):

/word-accessibility scan report.docx
/word-accessibility review the quarterly report template
/word-accessibility check all .docx files in /docs

In GitHub Copilot Chat:

@word-accessibility scan report.docx for accessibility issues
@word-accessibility check the employee handbook

Via the prompt picker: Select audit-single-document from the prompt picker and enter the .docx file path. This pre-configures the full strict scan profile and generates a DOCUMENT-ACCESSIBILITY-AUDIT.md report.

Via document-accessibility-wizard: For multi-document audits, the wizard orchestrates word-accessibility automatically - you do not invoke it directly.

Step-by-Step: What a Scan Session Looks Like

Here is a complete scan from start to finish so you know exactly what to expect.

You say:

/word-accessibility scan quarterly-report.docx

What the agent does:

1. Reads the DOCX file using the scan_office_document MCP tool, which parses the ZIP/XML structure of the .docx format and extracts the document body, styles, properties, and relationships.

2. Runs all 11 accessibility rules against the XML. For each rule:

   • Checks the relevant XML elements (e.g., every <a:blip> for alt text, <w:pStyle> for heading styles, <w:tr> for table header rows)
   • Determines whether a violation exists
   • Assigns a confidence level: High (definitely a violation), Medium (likely), Low (possible)

3. Computes the severity score using the weighted formula: errors = -10 points each (high confidence), warnings = -3 points each (high confidence), floor at 0.

4. Generates a structured report with findings grouped by severity. Here is what a real finding looks like:

DOCX-E001 [Error] - High Confidence
Image without alt text
Location: Paragraph 14, inline image
Remediation: Right-click the image in Word -> select "Edit Alt Text" -> describe what
the image shows. If the image is purely decorative (borders, dividers, backgrounds),
check "Mark as decorative" to suppress screen reader announcement instead.

5. Presents the score and grade, then offers next steps: generate a full VPAT report, run a folder audit, or hand off to the document-accessibility-wizard for deeper analysis.

Understanding Your Results

Score Interpretation

Score	Grade	What it means
90-100	A	Excellent - minor issues only, safe to distribute
75-89	B	Good - some warnings, fix before wide distribution
50-74	C	Needs work - multiple errors affecting AT users
25-49	D	Poor - significant barriers, remediate before sharing
0-24	F	Failing - likely unusable with screen readers

What to Fix First

Follow this order for maximum impact with minimum effort:

1. DOCX-E003 (No headings) - Affects all navigation. Screen reader users rely on the heading hierarchy to jump through long documents. Without headings, they must read linearly.
2. DOCX-E001 (Images without alt text) - Every image missing alt text is a content gap for blind users.
3. DOCX-E002 (Missing document title) - The document title is announced when the file opens. Missing it is a confusing first experience.
4. DOCX-E004/E005 (Tables without headers, missing language) - Affects table reading order and text-to-speech pronunciation.
5. DOCX-W002 (Skipped heading levels) - H1 -> H3 with no H2 confuses navigation; screen readers announce the level jump.

Confidence Levels

Not every finding is equally certain from static XML analysis alone:

• High confidence - The issue is definitively present in the XML (e.g., an image element with no alt text attribute).
• Medium confidence - The issue is likely present but context matters (e.g., a very long alt text string that may need summarization).
• Low confidence - A pattern that might be a problem depending on intent (e.g., empty paragraphs may be intentional spacing or may indicate a structural issue).

Focus your manual review effort on medium and low confidence items. High confidence findings can be addressed mechanically.

Connections

Connect to	When
document-accessibility-wizard	For full multi-document audits with cross-document scoring, VPAT generation, and CI/CD setup
office-scan-config	To configure which rules run and suppress rules that do not apply to your document type
excel-accessibility	For spreadsheets embedded in Word documents or when auditing a mixed Office collection
pdf-accessibility	Word documents are often exported to PDF - audit both if the PDF is distributed