Document Processing

Extract, convert, and structure content from documents. Load the project:document-processing skill for reference.

Process

1. Identify Input

!echo "Input: $ARGUMENTS"

Parse arguments:

• First argument: file path or directory
• --format: output format (markdown, json, text). Default: markdown

2. Detect Document Type

For each input file, determine processing approach:

• Check if PDF has native text layer
• Check if input is image (PNG, JPG, TIFF)
• Check if input is a directory (batch mode)

3. Process

Based on document type, use appropriate extraction method:

• Native PDF: pymupdf or pdftotext
• Scanned PDF/Images: Mistral OCR API or tesseract
• Word documents: python-docx or pandoc

4. Structure Output

Clean up extracted text:

• Fix OCR artifacts
• Reconstruct tables
• Identify headers
• Extract structured data if requested (emails, dates, etc.)

5. Save Results

Write output to file(s) in requested format alongside the input.