| name | youtube-transcript |
| description | Fetches the transcript (captions) from a YouTube video to use as context. Use this when the user provides a YouTube URL or video ID and wants to reference, summarise, or ask questions about the video's content. |
| argument-hint | <YouTube URL or video ID> |
YouTube Transcript
When the user provides a YouTube video URL or ID and wants to work with the video's content, fetch the transcript using the included script and use the output as context.
When to use this skill
- The user pastes a YouTube link and asks questions about it
- The user wants to summarise, translate, or extract information from a YouTube video
- The user references a YouTube video as background context for a task
Steps
-
Extract the video URL or ID from the user's message.
-
Run the fetch script to download the transcript:
node .github/skills/youtube-transcript/fetch-transcript.mjs "<YouTube URL or video ID>"
-
The script prints the transcript to stdout in this format:
# YouTube Transcript
**Title:** ...
**Channel:** ...
**Duration:** ...
**Language:** ...
## Transcript
<full transcript text>
-
Read the output and use the full transcript as context to answer the user's request.
Notes
- The script requires Node.js 18+ (uses built-in
fetch).
- It works with
youtube.com/watch?v=, youtu.be/, /embed/, /shorts/, or a raw 11-character video ID.
- If captions are unavailable (disabled by the uploader), the script exits with an error message — let the user know.
- Prefer English captions when available; the script falls back to the first available language.
Script reference
See fetch-transcript.mjs for the full implementation.