video-analysis

Analyze authorized video content from a local video file, a public direct HTTP(S) video URL, or—with --allow-platform—a single public video on a platform page (e.g. Bilibili/YouTube watch links, fetched via yt-dlp without login or cookies). Use this skill whenever the user asks to read, understand, summarize, transcribe, extract chapters or key moments from a video, or asks questions about a video’s contents—even if they refer to a video link rather than saying “video analysis.” It safely downloads direct media links, extracts audio and key frames, transcribes speech, optionally uses explicitly selected GPT, Grok, or any user-supplied OpenAI-compatible vision model (just an API key, endpoint, and model name) to interpret frames and GPT, Grok, DeepSeek, or custom text providers to synthesize a report, and produces timestamped results.

模型生成摘要（rules/v1 · 2026-09-21 17:19）：Analyze authorized video content from a local video file, a public direct HTTP(S) video URL, or—with --allow-platform—a …
这是模型对公开材料的总结，不是官方声明，请以原链内容为准。

存档时间线

版本	存档时间	内容哈希	内容
v1	2026-09-21 17:16	97bdbf39	可取

版本索引永久保留；内容副本只保留最近 2 版，更早版本仅留索引与哈希（存档时间线的证据链不会因此断裂）。

下载存档内容副本

纠错与举报（发现条目失效、署名有误或涉及侵权？）

提交举报 / 纠错

侵权举报经核验成立后，我们会即时下线该条目并删除已存的内容副本。