kuavi-analyze - SKILL.md Agent Skill

name: kuavi-analyze description: Full end-to-end video analysis with KUAVi agent: video-analyst context: fork argument-hint: disable-model-invocation: true

Perform comprehensive video analysis: index, search, reason, and answer.

Parse $ARGUMENTS for: video path and question/task.
Index: If the video is not yet indexed, call kuavi_index_video.
Orient: Call kuavi_get_scene_list to understand the video structure.
Search: Use the SEARCH-FIRST strategy to find relevant content:
- Decompose the question into search sub-queries
- Search across multiple fields (summary, action, visual, temporal)
- Search transcript for spoken content
Inspect: Extract frames for the most relevant segments.
Reason: Synthesize findings from visual evidence and transcript.
Verify: Cross-reference visual and audio evidence.
Answer: Provide a clear, evidence-based answer.

The first argument is the video path. Remaining text is the question.

Example: /kuavi-analyze /path/to/video.mp4 What is the main topic of this presentation?