YouTube → Readwise Article Pipeline

An automation pipeline that pulls the caption transcript from each video in a YouTube playlist, expands it into a polished, long-form article via the Gemini API using an outline-first prompt (to avoid the "middle-of-the-text" compression a single-pass summarization prompt produces on long videos), and pushes the result to Readwise Reader through their API — removing the video from the source playlist once it's safely saved, so nothing is processed twice. Runs automatically every hour via a GitHub Actions cron job, with credentials cached as repo secrets and a local OAuth token reused across runs rather than re-authenticating each time. A video with no usable transcript (disabled captions, none available) is skipped rather than processed with empty content.