View the full repo on GitHub → https://github.com/tga-cheetung/yt-digest
Paste a YouTube link. Get a vault-ready research digest in ~90 seconds for ~$0.04.
A Claude Code skill that turns any YouTube video into a single markdown note you can actually use: a TL;DR, 5–10 key takeaways each deep-linked to the exact second they happen, screenshots of the slides the speaker points at, and a collapsed full transcript at the bottom. The source video is downloaded, used for frame grabs, then deleted — your vault keeps ~3 MB per video, not 100.
It is not a content-repurposing factory. It does not turn one video into thirty social posts. It does the opposite: it turns a video into knowledge you keep — queryable, timestamped, and screenshotted in your second brain. Built for people who bookmark 40 videos a week and watch four.
vault/research/youtube/<slug>.md # the digest
vault/attachments/youtube/<slug>/ # frame-MM-SS.png screenshots
The digest has a fixed structure: frontmatter (source, channel, length, dates) so it's filterable later; a TL;DR of 3–5 standalone bullets; a Key Takeaways section where each numbered point carries a timestamp deep-link back to the source and, when the speaker references a visual, an embedded screenshot of that exact frame; an Open Questions section; and the full transcript collapsed at the bottom, every line timestamped.
A real example — a 47:55 interview — produced 8 TL;DR bullets, 10 numbered takeaways, and 6 embedded screenshots, generated in about 90 seconds for roughly four cents.
Four files, two local tools, no SaaS. yt-dlp pulls metadata, captions, and a 720p copy of the video. ffmpeg grabs single frames at the timestamps that matter. A short Python script cleans the transcript (including a rolling-merge pass that fixes YouTube's duplicate-heavy auto-captions). Claude does the judgment: it reads the full transcript and decides which 5–10 moments earn a bullet and which need a screenshot, guided by an editable selection brief.
Three knobs, all in SKILL.md: where digests land (default assumes Obsidian; change the paths to wherever your notes live), what counts as a key moment (the selection brief — point it at your domain), and whether to grab screenshots at all (--no-screenshots skips the video download for talking-head content).
/yt-digest <https://youtu.be/BI-MNjm1tTQ>
That's the whole interface. It also runs headless — the skill prints DIGEST: <path> as its last line, so any cron job, CI step, or chat bot can drive it. Wire it to a Slack bot and "drop a link, get a digest in your vault" becomes a one-message workflow from your phone.
GitHub: https://github.com/tga-cheetung/yt-digest (MIT licensed)
Created by The GTM Architects, 2026. Built for technical founders and lean GTM operators who'd rather keep knowledge than scroll past it.