Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
- Added a third-party content warning to the documentation, advising users to treat analysis results from YouTube and arbitrary URLs as untrusted data. - No functional changes. Documentation update only.
What This Skill Does
Analyzes video content using AI, accepting local files, YouTube URLs, or HTTP video URLs as input. Supports Google Gemini and Moonshot AI (Kimi) as backends. Returns summaries, answers to specific questions, and optional timestamped breakdowns.
Handles YouTube and HTTP URLs natively through Gemini without downloading, caches local uploads by content hash to skip re-uploads, and wraps multiple AI providers in one CLI instead of requiring separate tools per source type.
When to use it
- Summarizing a recorded team meeting without watching it
- Asking follow-up questions about a YouTube tutorial
- Getting a timestamped breakdown of a product demo video
- Identifying key moments in a dashcam or security clip
- Extracting spoken content or descriptions from an HTTP-hosted video
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Summarize the key moments in this YouTube video with timestamps
- 1Run `video-understand analyze "https://www.youtube.com/watch?v=VIDEO_ID" "What are the key moments?" --timestamps`
- 2Gemini receives the YouTube URL natively without downloading
- 3AI returns a timestamped breakdown of notable moments
- 4Agent formats and returns the summary to the user
A timestamped list of key moments from the video
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.