$npx -y skills add dkyazzentwatwa/chatgpt-skills --skill ocr-document-processorExtract text and structure from scans, images, and scanned PDFs. Use for OCR, searchable PDFs, table extraction, receipt parsing, and business card parsing.
| 1 | # OCR Document Processor |
| 2 | |
| 3 | Handle OCR-heavy inputs where text must be recovered from images or scanned pages. |
| 4 | |
| 5 | ## Use This For |
| 6 | |
| 7 | - OCR on images and scanned PDFs |
| 8 | - Searchable PDF export |
| 9 | - Structured extraction to text, markdown, JSON, or HTML |
| 10 | - Table extraction from scanned material |
| 11 | - Receipt parsing and business card parsing |
| 12 | |
| 13 | ## Workflow |
| 14 | |
| 15 | 1. Decide whether plain OCR, structured extraction, or document-specific parsing is needed. |
| 16 | 2. Preprocess noisy inputs before extraction when skew, blur, or shadows are present. |
| 17 | 3. Use `scripts/ocr_processor.py` for core OCR tasks. |
| 18 | 4. Use the focused helpers when the input is specialized: |
| 19 | - `scripts/business_card_scanner.py` |
| 20 | - `scripts/receipt_scanner.py` |
| 21 | 5. Return confidence caveats when the source is low quality, rotated, handwritten, or multilingual. |
| 22 | |
| 23 | ## Guardrails |
| 24 | |
| 25 | - Prefer explicit language selection when accuracy matters. |
| 26 | - Do not claim fields are exact when OCR confidence is weak. |
| 27 | - Route non-scanned digital PDFs to `document-converter-suite` instead of OCR by default. |