2.6k
GitHub Stars
4
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill yt-summary- _meta.json280 B
- README.md3.8 KB
- requirements.txt16 B
- SKILL.md4.5 KB
Overview
This skill summarizes any YouTube video when you drop the link in chat. It extracts the transcript via TranscriptAPI.com, then generates a structured summary that includes a TL;DR, key points, and notable quotes when available. You can also append custom instructions after the URL to tailor the summary focus.
How this skill works
The skill calls TranscriptAPI to fetch the video's captions and parses a structured JSON result that includes header, transcript text, language, token count, title, channel, and duration. It refuses to fabricate missing content and stops with a clear error if transcript extraction fails. For short transcripts it summarizes in one pass; for very long transcripts it summarizes the first ~40K tokens and notes the truncation. If you provide extra text after the URL, those words are treated as a custom prompt appended to the summarization instructions.
When to use it
- You need a concise, structured summary of a video without watching the whole thing
- You want summaries focused on a specific aspect (technical details, arguments, timestamps, etc.) by adding a short prompt
- You have a video with captions or auto-generated transcript available
- You want quick highlights and quotable lines for notes, research, or sharing
- You need output kept under messaging limits (e.g., Telegram 4000 chars)
Best practices
- Paste only the YouTube URL, optionally followed by a short instruction (e.g., 'focus on methodology')
- Use the secure API key storage option if running extraction locally to avoid exposing keys
- Expect an explicit error if the transcript is unavailable or extraction fails—do not assume content was created
- When requesting deep analysis, include precise guidance in the custom prompt to shape the summary
- If the video is extremely long, accept that only the first ~40K tokens will be summarized and ask for follow-ups if needed
Example use cases
- Researcher: Get a quick TL;DR and key points from a 90-minute conference talk to decide relevance
- Student: Paste a lecture link and request a focus on formulas and definitions
- Journalist: Extract notable quotes and succinctly summarize an interview
- Developer: Ask for a summary that highlights technical implementation details from a tutorial
- Content curator: Generate shareable bullets and a short quote to accompany a link post
FAQ
The skill will report that no transcript is available and stop. It does not fabricate content or fall back to other web searches.
How do I request a specific focus in the summary?
Append a short instruction after the URL in the same message (e.g., 'focus on the experimental setup') and the skill will include it when generating the summary.
What happens for very long videos?
If the transcript exceeds a large token threshold, the skill summarizes roughly the first ~40K tokens and clearly notes the truncation so you can request additional segments if needed.