Agent: PDF Accessibility

pdf-accessibility - PDF Document Accessibility

Scans PDF documents for conformance with PDF/UA (ISO 14289) and the Matterhorn Protocol. Uses the scan_pdf_document MCP tool to parse PDF files and check tagged structure, metadata (title, language), bookmarks, form field labels, figure alt text, table structure, font embedding, and encryption restrictions.

When to Use It

• Reviewing PDFs before publishing or distributing
• Checking PDF conformance for procurement (Section 508, EN 301 549)
• Auditing scanned documents for basic structural accessibility
• Verifying PDF/UA compliance after conversion from Office documents

Rule Layers

Expand - 3 rule layers (56 rules total)

Layer	Rules	Purpose
PDFUA.*	30 rules	PDF/UA conformance - tagged structure, metadata, navigation, forms, tables, fonts
PDFBP.*	22 rules	Best practices beyond PDF/UA requirements
PDFQ.*	4 rules	Pipeline quality - file size limits, scan detection, encryption checks

Key Checks

Expand - 10 key PDF/UA checks

• Missing tagged structure (PDFUA.TAGS.001)
• No document title in metadata (PDFUA.META.001)
• Missing document language (PDFUA.META.002)
• Figures without alt text (PDFUA.TAGS.004)
• Tables without headers (PDFUA.TAGS.005)
• Unlabeled form fields (PDFUA.FORM.001)
• Missing bookmarks (PDFUA.NAV.001)
• Non-embedded fonts (PDFUA.FONT.001)
• Scanned image PDFs (PDFQ.SCAN.001)
• Encryption restricting assistive technology (PDFQ.ENC.001)

Example Prompts

Show example prompts

/pdf-accessibility scan contract.pdf for PDF/UA compliance
@pdf-accessibility review the annual report PDF
@pdf-accessibility check all PDFs in the legal/ directory
@pdf-accessibility what PDFUA rules does this file violate?

How to Launch It

In Claude Code (terminal):

/pdf-accessibility scan annual-report.pdf
/pdf-accessibility check all PDFs in /legal for PDF/UA compliance
/pdf-accessibility what PDFUA rules does brochure.pdf violate?

In GitHub Copilot Chat:

@pdf-accessibility scan contract.pdf
@pdf-accessibility check the procurement RFP for Section 508 compliance

Via the prompt picker: Select audit-single-document and enter the .pdf path. For government/procurement PDFs, consider using generate-vpat afterwards to produce a formal ACR.

Via document-accessibility-wizard: For bulk PDF audits or mixed-format collections, the wizard handles PDF scanning alongside Office document scanning.

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

You say:

/pdf-accessibility scan annual-report.pdf

What the agent does:

1. Reads the PDF using the scan_pdf_document MCP tool, which parses the PDF's structure tree, metadata dictionary, content streams, and form definitions.

2. Runs all applicable rules from the three layers. The order matters:

   • PDFQ rules run first (pipeline quality - is this a scanned image PDF? Is it encrypted in a way that blocks AT?)
   • PDFUA rules run next (structural conformance - tags, metadata, navigation, forms, fonts)
   • PDFBP rules run last (best practices beyond the standard)

3. Assigns confidence levels. PDF/UA conformance checking is more nuanced than Office checking because some rules (e.g., reading order correctness) require human judgment.

4. Returns findings with PDF object identifiers. Here is a real example:

PDFUA.TAGS.004 [Error] - High Confidence
Figure without alt text
Location: Page 4, Figure object #142
Remediation: Open in Adobe Acrobat Pro -> Tools -> Accessibility -> Reading Order ->
click the figure -> add alternative text in the Alt Text field. In the source (Word/InDesign),
add alt text before PDF export to avoid needing to retrofit.

5. Delivers the score, grade, rule violation list, and next steps.

Understanding Your Results

Score Interpretation

Score	Grade	What it means
90-100	A	Excellent - near PDF/UA conformance
75-89	B	Good - minor gaps, addressable with Acrobat Pro
50-74	C	Needs work - multiple structural barriers
25-49	D	Poor - significant AT barriers, remediation required
0-24	F	Failing - untagged or scanned-image PDF, essentially inaccessible

The Verdict on Scanned PDFs

PDFQ.SCAN.001 triggers when a PDF consists primarily of scanned images with no selectable text. This is the worst possible accessibility outcome for a PDF - a screen reader cannot read a scanned image. The only remediation is OCR (Optical Character Recognition) to convert the scanned pages to selectable text, followed by a full tagging pass. The agent will flag this immediately if detected and recommend stopping the audit until remediation is complete.

What to Fix First

1. PDFQ.SCAN.001 (Scanned image PDF) - If this fires, nothing else matters until OCR is complete.
2. PDFQ.ENC.001 (Encryption blocking AT) - If the PDF's security settings block screen readers, fix the permissions before proceeding.
3. PDFUA.TAGS.001 (Missing tagged structure) - An untagged PDF is not navigable. All content must be tagged.
4. PDFUA.META.001/002 (Missing title and language) - Both are required for PDF/UA conformance and are trivially fixed.
5. PDFUA.FORM.001 (Unlabeled form fields) - If the PDF is a form, every field must have an accessible name.

PDF Remediation Reality

PDF remediation is more involved than Office document remediation because PDFs are a presentation format, not an authoring format. The best practice is to fix accessibility issues in the source document (Word, InDesign, PowerPoint) and re-export. Retrofitting a complex PDF in Acrobat Pro is time-consuming and error-prone. Use pdf-accessibility to identify issues; use the source document agents to fix them at the root.

Connections

Connect to	When
document-accessibility-wizard	For batch PDF audits, cross-document pattern analysis, and VPAT generation
pdf-scan-config	To configure rule layers - e.g., run PDFUA only (skip PDFBP best practices) for a procurement baseline
word-accessibility	When the PDF was generated from Word - fix issues at the source before re-exporting
powerpoint-accessibility	When the PDF was generated from PowerPoint - same source-first remediation approach