Video Understanding

Analyze and summarize videos from 1000+ sites using Google Gemini AI, providing transcripts, descriptions, summaries, and answers to questions.

This skill appears to do what it says: it downloads videos (via yt-dlp), may remux/merge with ffmpeg, and uploads the content to Google Gemini using the GEMINI_API_KEY for analysis. Before installing, consider: 1) Privacy and copyright — uploaded videos will be sent to Google's servers, so avoid private/confidential or copyrighted content unless you have rights. 2) Billing and limits — using Gemini File API may incur costs and has size limits; confirm your API plan. 3) Trusting yt-dlp downloads — yt-dlp executes network downloads and may write files locally; ensure you trust the video sources. 4) Runtime dependencies — brew will install system binaries and the Python google-genai package will be installed; review these if you have strict policy controls. If these behaviors are acceptable, the skill is internally consistent.

Type: OpenClaw Skill Name: video-understanding Version: 1.1.0 The skill bundle is classified as benign. It transparently describes its purpose to analyze videos using Google Gemini and `yt-dlp`. The `SKILL.md` contains no prompt injection attempts against the OpenClaw agent. The `analyze_video.py` script correctly uses `subprocess.run` with a list of arguments to invoke `yt-dlp`, preventing shell injection vulnerabilities from user-provided URLs. It handles temporary files and API keys (GEMINI_API_KEY) as expected for its stated functionality, without evidence of data exfiltration, persistence, or other malicious behaviors.