youtube-transcript
Installation
SKILL.md
YouTube Transcript & Frame Extractor
This skill exists to stop: quoting video content that can't be traced — a quote without video-id + timestamp is a [GUESS] borrowing the video's authority.
🤖 0. OUTPUT RULE
Every transcript quote carries video-id + timestamp (or .txt line number); downloaded transcripts are saved into the repo's transcripts folder so others can verify later.
Two jobs: transcript → get_transcript.py or browser_capture.js (§1–2) · slide frames → capture_frames.py (§4).
Transcript: two paths — pick correctly upfront
| Path A: python script | Path B: browser | |
|---|---|---|
| File | get_transcript.py |
browser_capture.js |
| Runs on | Personal machine (residential IP) | Chrome on a personal machine |
| Speed | Fast, one command | Slower, 3 steps |
| Fails when | Cloud IPs are blocked by YouTube → IpBlocked, exit code 2 |
Almost never |
⚠️ Inside containers / CI / cloud sandboxes, path A and
capture_frames.pyboth fail (YouTube blocks cloud IP ranges). The script exits 2 and prints the fallback instructions. Don't proxy around it — use path B.