← 返回 Skills 市场
yilsonyan

video-summarize

作者 yilsonyan · GitHub ↗ · v1.0.2026041402 · MIT-0
cross-platform ⚠ suspicious
87
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install video-summarize
功能描述
视频总结,触发:1.用户给视频链接(B站/YouTube/抖音/Twitter/TikTok等)、2.总结这个视频
安全使用建议
This skill largely implements video→subtitle→transcribe→summarize, but it performs actions you should review before installing: (1) install_dependency.sh will attempt to install Homebrew (via curl of the official installer), add brew to your shell profile (~/.zprofile), install system packages (ffmpeg, whisper-cpp) and pip packages in a virtualenv, and download a Whisper model file; (2) process.sh will use yt-dlp with --cookies-from-browser chrome which accesses your browser cookies (potentially exposing authenticated content and private cookies); (3) the scripts assume a Homebrew-capable environment (macOS/Linux with brew) despite no OS restriction. Recommendations: inspect the install and process scripts line-by-line; run the install steps manually rather than as a single unattended script; if you don't want the skill to access browser cookies, remove/avoid the --cookies-from-browser option; consider running the skill in an isolated VM/container; and if you are uncomfortable with Homebrew or shell-profile modifications, do not run the installer or adapt it to your environment (install dependencies yourself and supply the model). If you need a lower-privilege alternative, ask the author to provide a containerized image or clearer, non-invasive install instructions.
功能分析
Type: OpenClaw Skill Name: video-summarize Version: 1.0.2026041402 The skill automates video transcription and summarization using yt-dlp and whisper.cpp, but it includes high-risk behaviors. Specifically, scripts/process.sh attempts to access sensitive browser data by using the '--cookies-from-browser chrome' flag when processing YouTube links, which could expose session data. Additionally, scripts/install_dependency.sh uses a curl|bash pattern to install Homebrew and performs system-level package installations, which are high-privilege actions that could be abused if the source is compromised.
能力评估
Purpose & Capability
The declared purpose (video summarization using subtitles or Whisper) matches the scripts: yt-dlp, ffmpeg, and whisper.cpp are used as expected. However the script uses yt-dlp's --cookies-from-browser chrome for YouTube links, which implies reading browser cookies (broader access than 'public videos' claim) and the installer assumes Homebrew is available/appropriate despite no OS restriction in metadata.
Instruction Scope
Runtime instructions and scripts instruct the agent/user to run install_dependency.sh and process.sh. process.sh invokes yt-dlp with --cookies-from-browser chrome (accesses browser cookies), reads/writes filesystem under the skill directory and ~/.zprofile may be modified by the installer. The SKILL.md promises only public videos, but use of browser cookies can access authenticated/region-restricted content — this is scope creep and a privacy risk.
Install Mechanism
The install script will curl and run the Homebrew installer (raw.githubusercontent.com), install brew packages, write to ~/.zprofile for brew environment, create a virtualenv and pip-install yt-dlp, and curl a model file from a Hugging Face URL. These are traceable sources but are system-changing actions (installer execution, shell profile modification, package installs) and assume a Homebrew-capable environment even though the skill metadata declares no OS restrictions.
Credentials
No environment variables or config paths are declared, yet the code implicitly accesses browser cookies via yt-dlp (sensitive data) and will alter the user's shell profile to add Homebrew to PATH. Required privileges (ability to install Homebrew and brew packages, modify ~/.zprofile) are not surfaced in the skill metadata.
Persistence & Privilege
The skill is not always-enabled, but the installer makes persistent changes: it can install Homebrew system-wide and append to ~/.zprofile, create a .venv and download models into the skill directory. These persistent system changes increase blast radius if you install without review.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install video-summarize
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /video-summarize 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.2026041402
- Updated dependency installation to explicitly require Python 3 and explain Python virtual environment use. - Revised "Dependency Installation" section, now includes Python 3 installation instructions and command. - Moved and streamlined "Features" section for better readability. - Clarified automated deletion of temporary cache files after processing. - No functional changes to processing logic or workflow.
v1.0.20260414
视频总结 v1.0.20260414 – 重大更新:支持智能字幕优先 - 新增自动优先提取官方/人工字幕,无字幕时才用 Whisper 音频转录,大幅提升转录准确率与速度 - 增强多平台兼容性,字幕支持 YouTube/B站等主流平台(优先中文人工字幕) - 优化并发处理,每个视频独立缓存与临时目录,更加稳定可靠 - 处理完成后自动清理临时文件,节省磁盘空间 - 更简洁的流程说明与用法文档
v1.0.0
Initial release: Summarize online videos from 1000+ platforms (Bilibili, YouTube, Douyin, Twitter, TikTok, etc.) via link input. - Supports multi-language transcription using Whisper; automatic language detection. - Async processing with status prompts and smart caching—repeat requests are instant. - Audio is extracted, transcribed, and summarized; results stored as .txt and .md files with safe auto-renamed filenames. - Installation guides and script automation for all required dependencies. - Not limited to a single platform—compatible with any service supported by yt-dlp.
元数据
Slug video-summarize
版本 1.0.2026041402
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 3
常见问题

video-summarize 是什么?

视频总结,触发:1.用户给视频链接(B站/YouTube/抖音/Twitter/TikTok等)、2.总结这个视频. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 87 次。

如何安装 video-summarize?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install video-summarize」即可一键安装,无需额外配置。

video-summarize 是免费的吗?

是的,video-summarize 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

video-summarize 支持哪些平台?

video-summarize 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 video-summarize?

由 yilsonyan(@yilsonyan)开发并维护,当前版本 v1.0.2026041402。

💬 留言讨论