← Back to Skills Marketplace
vsum
by
Chrischaan
· GitHub ↗
· v1.0.0
734
Downloads
3
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install vsum
Description
视频 summarizer,支持 YouTube 和 Bilibili 视频自动获取字幕并 AI 总结,输出为 md 格式。适用于:用户给出一个视频链接,希望总结内容。
Usage Guidance
This skill appears to do what it says (download subtitles with yt-dlp and summarize them with an AI), but there are privacy and transparency gaps you should consider before installing:
- You must have yt-dlp installed separately; the skill metadata doesn't state that as a required binary.
- For Bilibili it uses yt-dlp's --cookies-from-browser option, which reads your browser's cookies (session tokens). That can expose sensitive session data; only run this if you trust the environment and understand the cookie access.
- The SKILL.md mentions using external AI providers but does not declare how API keys are provided. Expect to supply an API key/credential to the agent; ensure you use a trusted provider and understand that subtitle text will be transmitted to that provider.
- The helper script writes subtitle files to ~/Downloads; check that location if you are concerned about local storage or accidental disclosure.
- If you need stronger guarantees, ask the author to: (1) add yt-dlp and required API credentials to the skill metadata, (2) document exactly where summaries are sent and how credentials are stored/used, and (3) avoid or clearly warn about reading browser cookies (or provide alternate auth flows).
Capability Analysis
Type: OpenClaw Skill
Name: vsum
Version: 1.0.0
The skill is classified as suspicious due to its use of `yt-dlp` with the `--cookies-from-browser chrome` option in `scripts/vsum.sh` and described in `SKILL.md`. While this functionality is stated as necessary for accessing Bilibili content, it grants the script access to potentially sensitive browser cookies. This capability, even if justified by the skill's purpose, represents a significant security risk and a potential vector for sensitive data exposure if the agent or skill were compromised, elevating it beyond benign.
Capability Assessment
Purpose & Capability
The name/description (video summarizer for YouTube/Bilibili) aligns with the provided script and instructions: yt-dlp is used to fetch subtitles and then an AI API is used to summarize. However the skill metadata does not declare required tooling (yt-dlp) or the need for an AI API key, which are necessary for normal operation.
Instruction Scope
SKILL.md and scripts instruct use of yt-dlp including the flag --cookies-from-browser chrome for Bilibili, which reads browser session cookies; this is a sensitive operation but is not called out in requires.config or explained/limited. The SKILL.md also instructs agents to ‘send subtitle content to AI’ without specifying destination handling, retention, or how API keys are supplied, granting broad discretion to transmit user data to external providers.
Install Mechanism
Instruction-only skill with a small helper script and no install spec. This minimizes on-disk installs; risk comes from external tools (yt-dlp) and network activity rather than an installer pulling arbitrary code.
Credentials
requires.env lists none, yet the runtime clearly needs: (1) an AI provider API key/credentials to perform summaries, and (2) read access to browser cookies for Bilibili subtitle downloads. Not declaring these is a mismatch — API keys and cookie access are sensitive and should be explicitly required and justified in metadata.
Persistence & Privilege
The skill does not request always:true or modify system/other-skill configs. It runs on-demand and writes subtitle files to the user's Downloads directory, which is reasonable for its purpose but should be made explicit to users.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install vsum - After installation, invoke the skill by name or use
/vsum - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release - YouTube & Bilibili video summarizer
Metadata
Frequently Asked Questions
What is vsum?
视频 summarizer,支持 YouTube 和 Bilibili 视频自动获取字幕并 AI 总结,输出为 md 格式。适用于:用户给出一个视频链接,希望总结内容。 It is an AI Agent Skill for Claude Code / OpenClaw, with 734 downloads so far.
How do I install vsum?
Run "/install vsum" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is vsum free?
Yes, vsum is completely free (open-source). You can download, install and use it at no cost.
Which platforms does vsum support?
vsum is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created vsum?
It is built and maintained by Chrischaan (@chrischaan); the current version is v1.0.0.
More Skills