菜单
Use when the user provides a video or audio URL and asks to summarize, extract subtitles, create video notes, 视频总结, 视频笔记, or summarize content from Bilibili, YouTube, Douyin, Xiaohongshu, TikTok, or yt-dlp supported sites.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Socratic decision and idea-clarification coach. Use when the user asks to think through a decision, clarify an idea, evaluate what they really want, explore why they want it, define success criteria, find fake goals, identify constraints, detect goal drift, choose the key goal among multiple goals, or decide the next best step. Trigger on English or Chinese requests such as "thinking partner", "帮我梳理思路", "帮我想清楚", "我该怎么选", "这个 idea 怎么想清楚", "why", "目标是什么", or "苏格拉底教练". Do not use for requests that already have a clearly specified implementation task unless the user explicitly asks to clarify goals first.
Use when working in this repository on industry research, raw source organization, knowledge synthesis, investment analysis, career analysis, industry templates, or reusable research tooling.
| name | video-summarizer |
| description | Use when the user provides a video or audio URL and asks to summarize, extract subtitles, create video notes, 视频总结, 视频笔记, or summarize content from Bilibili, YouTube, Douyin, Xiaohongshu, TikTok, or yt-dlp supported sites. |
Extract the transcript first, then summarize from evidence. Do not invent content when extraction fails.
This skill is a Codex-native port of keepongo/video-summarizer at upstream commit 481a772.
bilibili.com/video/, b23.tv/, BV...youtube.com/watch, youtu.be/, youtube.com/shorts/douyin.com, v.douyin.comxiaohongshu.com, xhslink.comyt-dlpreferences/dependencies.md.scripts/video_subtitle.py with uv run python ...; add uv --group ... dependency groups only when the platform needs them.error, report the failure and the next concrete fix.subtitle_text. If frames are present, embed selected images with absolute paths.references/output-format.md.Bilibili or direct-platform extraction:
uv run python .agents/skills/video-summarizer/scripts/video_subtitle.py "<VIDEO_URL>"
YouTube:
uv run --group video-youtube python .agents/skills/video-summarizer/scripts/video_subtitle.py "<VIDEO_URL>"
TikTok, generic sites, or yt-dlp fallback:
uv run --group video-ytdlp python .agents/skills/video-summarizer/scripts/video_subtitle.py "<VIDEO_URL>"
OpenAI Whisper API fallback:
OPENAI_API_KEY=... uv run --group video-ytdlp --group video-whisper-api python .agents/skills/video-summarizer/scripts/video_subtitle.py "<VIDEO_URL>"
Local Whisper fallback is heavy. Use it only when the user explicitly wants audio transcription and accepts model downloads:
uv run --group video-ytdlp --group video-whisper-local python .agents/skills/video-summarizer/scripts/video_subtitle.py "<VIDEO_URL>"
Clear cached transcript and screenshots:
uv run python .agents/skills/video-summarizer/scripts/video_subtitle.py --clear-cache
references/config.example.json to .agents/skills/video-summarizer/config.json for local settings.references/config.whisper-api.example.json or references/config.whisper-local.example.json when enabling Whisper.references/setup.md when initializing this skill on a new machine or checking environment readiness.config.json, cookies, cache, and screenshots untracked.OPENAI_API_KEY for OpenAI Whisper API instead of writing a key into config.uv run --group ... command.references/dependencies.md.