youtube-watcher

What This Skill Does

The youtube-watcher skill is a specialized tool for the OpenClaw agent designed to bridge the gap between multimedia content and text-based analysis. By integrating with yt-dlp, this skill can programmatically fetch the closed captions or auto-generated transcripts of any publicly available YouTube video. Once the transcript is retrieved, it is converted into a text format that the agent can read, parse, and process. This transforms unstructured video content into searchable, summarizeable data, allowing users to extract key insights without needing to watch the entire video duration. It essentially turns your AI agent into a video librarian capable of recalling specific points, explaining concepts mentioned in lectures, or providing concise summaries of long-form interviews.

Installation

To integrate this capability into your OpenClaw environment, you must ensure that your system has the required dependency installed. First, confirm that the yt-dlp library is accessible in your system's PATH. Once confirmed, you can add the skill to your agent by executing the following command in your terminal: clawhub install openclaw/skills/skills/naliehu/youtube-watcherkx. After successful installation, the agent will gain the ability to call the get_transcript.py script automatically whenever a relevant YouTube link is processed.

Use Cases

This skill is highly versatile for users who consume a large volume of video content. Common use cases include:

• Academic Research: Summarizing educational videos or tutorials to quickly capture the core teaching points.
• Content Creation: Extracting quotes or specific information from video podcasts or news reports for use in blog posts or documentation.
• Meeting Analysis: Reviewing recorded meetings uploaded to YouTube to extract action items, decisions, and deadlines discussed.
• Technical Learning: Rapidly checking if a programming tutorial covers a specific error or library function before committing to watching the full video.

Example Prompts

1. "Watch this YouTube video at https://www.youtube.com/watch?v=dQw4w9WgXcQ and provide me with a bulleted summary of the main arguments made."
2. "I need to know if the speaker in this video mentions 'distributed systems' at any point. Can you check the transcript? https://www.youtube.com/watch?v=dQw4w9WgXcQ"
3. "Please extract the key steps for setting up a development environment from this tutorial: https://www.youtube.com/watch?v=dQw4w9WgXcQ"

Tips & Limitations

• Subtitle Dependency: This skill relies entirely on existing subtitles or closed captions. If a video has no transcript available, the script will return an error.
• Performance: For extremely long videos, the transcript text can be very large. Be mindful of your agent's context window limits when processing multi-hour videos.
• Accuracy: The accuracy of the transcript is directly tied to the quality of the YouTube-generated captions; automated captions may contain errors in technical jargon or non-native accents.
• Updates: Ensure yt-dlp remains updated, as YouTube frequently changes its front-end structure which can temporarily break retrieval scripts.