mediar-ai/packages/skills/skills/pdf-data-extractor

PDF Data Extractor

Extract structured data from PDF documents using OCR and pattern matching.

Prerequisites

• Chrome browser installed
• PDF file accessible locally or via URL

Instructions

1. Open Chrome and navigate to the PDF file:

   • If local file: use file:///path/to/document.pdf
   • If URL: navigate directly to the PDF URL

2. Wait for the PDF to fully load in Chrome's PDF viewer:

   • Confirm the page count is visible
   • Confirm text is selectable (if not scanned)

3. Use Chrome DevTools to extract text content:

   • Press F12 to open DevTools
   • Go to Console tab
   • For each page, extract visible text

4. For scanned PDFs without selectable text:

   • Take a screenshot of each page
   • Use OCR to extract text from screenshots
   • Clean up OCR results for accuracy

5. Parse extracted text for structured data:

   • Look for key-value patterns (e.g., "Name: John Doe")
   • Identify tables by detecting aligned columns
   • Extract dates, amounts, and other formatted data

6. Output the extracted data in requested format:

   • JSON for structured data
   • CSV for tabular data
   • Plain text for unstructured content

Error Handling

• If PDF is password protected, inform user and request password
• If OCR quality is poor, suggest higher resolution scan
• If PDF has complex layouts, process page by page manually

Notes

• Works best with text-based PDFs
• Scanned documents require OCR and may have lower accuracy
• For invoices and receipts, consider the invoice-parser skill