What it does
Opens each PDF in the media library and checks for a real, searchable text layer versus a scanned image with no OCR, correct heading and tag structure, a logical reading order, and alt text on any embedded images. Produces a per-file pass/fail report so the user knows which PDFs need to be re-exported or remediated rather than just replaced with an unhelpful warning.
Install
This skill isn't packaged in an installable registry. It's a plain SKILL.md file you can copy directly. Copy it from the SKILL.md tab and save it at the path for your assistant:
Use it in your AI assistant
Claude Code
.claude/skills/pdf-accessibility-checker/ (or ~/.claude/skills/pdf-accessibility-checker/ for all projects)
GitHub Copilot (VS Code)
.github/skills/pdf-accessibility-checker/
OpenAI Codex
.codex/skills/pdf-accessibility-checker/
Cursor
.cursor/skills/pdf-accessibility-checker/ (or ~/.cursor/skills/pdf-accessibility-checker/ globally)
Compatibility
Requires the agent to open and parse actual PDF file contents and structure, not just filename or upload metadata.
The complete SKILL.md for this skill:
---
name: pdf-accessibility-checker
description: Use when a user wants PDFs on their site checked for accessibility, mentions tagged PDFs, screen-reader-readable documents, or PDF compliance.
---
# PDF Accessibility Checker
## When to use
The site has uploaded PDFs (guides, reports, forms) and the user wants to know whether they're accessible to screen reader users, or is auditing before publishing new ones.
## Procedure
1. List the PDFs in the media library, or a specific set the user points to.
2. Open each file and check whether it has a real text layer (selectable, searchable text) versus being a flat scanned image.
3. Check for a logical tag structure: headings in order, lists marked as lists, tables with header cells identified.
4. Check that images embedded in the PDF carry alt text, and that the reading order matches the visual layout (common failure in multi-column PDFs).
5. Produce a per-file report: pass, or specific failures found, without attempting to silently "fix" the PDF itself.
6. For scanned-image-only PDFs, flag that OCR and manual tagging (usually outside the agent's own reach) are needed before the file can be considered accessible.
## Guardrails
- Don't claim a PDF is accessible based on filename or page count alone, always open and inspect the actual structure.
- Remediating PDF tag structure typically requires dedicated PDF tools, flag files that need that rather than attempting an in-place rewrite.