reMarkable notebooks: fetch, transcribe, summarize

Handwritten pages are not text anywhere in the pipeline. The tablet stores strokes, exports strokes, and the PDF contains strokes. The OCR step is your own vision — you read rendered page images. There is no OCR binary in this skill and none is wanted: handwriting is exactly the case where a general vision model beats Tesseract and friends.

The pipeline:

tablet ──► PDF ──► per-page PNG ──► you read them ──► transcript.md + summary.md
        (1)     (2)              (3)                (4)

1. Get the document list