Route visual questions about video clips to the VSS video_understanding tool.
npx skills add https://github.com/rblake2320/vigil --skill vss-ask-video-rblake2320
Please help me install this Agent Skill. Skill: vss-ask-video Source: https://github.com/rblake2320/vigil/tree/main/.claude/skills/vss-ask-video Command: npx skills add https://github.com/rblake2320/vigil --skill vss-ask-video-rblake2320
This skill enables users to obtain fresh, frame-level insights about a recorded video by querying the VSS agent's video_understanding tool, especially when existing outputs or metadata cannot answer.
Provide a fresh visual question to the VSS agent for the specified sensor/clip and return the analysis.
High-intent search queries and answers about installing and using this skill.
Visual question answering on video clips works by routing queries to the VSS video_understanding tool, which runs fresh frame analysis on a named sensor to extract visual facts like objects, actions, and safety details.
Yes, you can run ad-hoc video understanding queries on a named sensor without searching prior tool outputs, because the skill enforces guardrails to ensure fresh frame analysis is performed for each visual question.
Yes, enabling visual question answering requires a VSS base or lvs profile with the video_understanding tool exposed on the agent, plus a list of sensors available via the VST API.
VSS frame analysis retrieves visual facts about objects, people, actions, colors, timing, and safety details from video clips by running fresh frame-level analysis on the specified sensor.
You should avoid using prior tool outputs for video clip visual analysis whenever a new frame analysis is needed, because the skill enforces guardrails that require fresh queries to the VLM for accurate visual facts.
To start querying fresh visual insights, provide a visual question to the VSS agent for the specified sensor or clip, and the agent will return the analysis by running the video_understanding tool on the requested frames.
sayalinvidiaAnswer visual questions about recorded video clips using VSS video_understanding.
bettercallfanTracks detected objects across video frames and outputs Track schema with stable IDs, trajectories, movement states.
rblake2320Generate structured video analysis reports from VSS clip and incident data.
sayalinvidiaGenerate structured video analysis reports from VSS clips and incidents.
