Local Document Operations
Use local deterministic tools before spending model context.
Preferred order:
- Use structured parsers for PDFs and office documents.
- Use Apple Vision OCR for images on macOS.
- Use transcription for audio or video.
- Only then ask a model to interpret unclear content.
Keep raw extracted text in cache or inbox. Write only refined knowledge into the vault.
Bundled helper:
python skills/local-doc-ops/pdf-kit.py --help