Video Expert Analyzer

What it does

Video Expert Analyzer applies AI-driven visual analysis to identify, score, and extract high-quality scenes from videos. It runs a pipeline to download video, detect scenes, extract frames, and then scores each scene across five dimensions (aesthetic beauty, credibility, impact, memorability, fun/interest). The tool can operate in Agent mode (multimodal model views frames directly) or API mode (remote vision model). Outputs include scene_scores.json, per-scene reports, and a best_shots folder with selected clips.

When to use it

Use this skill when you need professional-grade curation: highlight reels, social clips, competitor analysis, or batch-quality checks. Trigger when the user asks for video analysis, scene scoring, extracting highlights, or mentions platforms like Bilibili/YouTube/Douyin/Xiaohongshu.

What's included

• Scripts: pipeline_enhanced.py, ai_analyzer.py, download helpers (described in SKILL.md) — (note: has_scripts=false in fetch output but the body documents CLI scripts)
• References: in-body methodology and prompt templates for Agent/API modes
• Instructions: step-by-step pipeline (download → scene detect → frame extract → AI scoring → reporting) and batching rules for large sets of scenes

Compatible agents

Infers compatibility with multimodal-capable agents and CLIs: Gemini (visual), Kimi, Claude (visual variants). Agent-mode works where the host assistant can view images; API mode requires a visual-model endpoint.