← 返回 Skills 市场
video-content-watcher
作者
Franklinyung
· GitHub ↗
· v1.0.0
· MIT-0
124
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install video-content-watcher
功能描述
Extract core content and generate structured analysis reports from YouTube, Bilibili, or local video files. Triggers when user asks to summarize, analyze, ex...
安全使用建议
Do not install this skill without addressing the mismatches. Specific concerns:
- Registry vs runtime mismatch: The registry declares no required env vars or binaries, but SKILL.md requires MINIMAX_API_KEY (required) and optionally WHISPER_API_KEY plus system tools (yt-dlp, ffmpeg). Expect to supply LLM/Whisper credentials and install ffmpeg/yt-dlp or the skill will fail.
- Hardcoded dev paths / external code execution: The small bundled scripts purposely insert an absolute/workspace-relative path into sys.path and import video_reader_mcp from workspace-code-dev/video-reader/src — that 'src' is NOT bundled. At runtime the skill will execute whatever Python module is found at that path on the host. This can lead to arbitrary code execution if the path points to untrusted code.
- Missing implementation: The real implementation appears to live outside the package. Ask the author for the full 'src' or a packaged module. Without it, the behavior depends entirely on host files and is unsafe to run on sensitive systems.
- Suggested mitigation before use: require the publisher to (a) include the src implementation in the skill bundle or provide a canonical, verified install path; (b) remove hardcoded /home/ykl paths and use relative imports or configurable paths; (c) update registry metadata to declare required env vars and needed system binaries; (d) review the full VideoReaderMCP code for network endpoints, credential use, and data exfiltration; (e) run the skill in a sandboxed environment and avoid providing real API keys until code is audited.
- What would change my assessment: If the publisher supplies the missing src files within the package (so no host-path imports are needed), removes hardcoded paths, and updates the registry to accurately list required env vars/binaries, this would make the package much more coherent and lower-risk.
能力评估
Purpose & Capability
The declared purpose (extract audio, transcribe, analyze with MiniMax) reasonably requires an LLM API key and possibly Whisper API/local model and system binaries (yt-dlp, ffmpeg). The registry metadata, however, lists no required env vars or binaries. That mismatch (missing MINIMAX_API_KEY and optional WHISPER_API_KEY, and no mention of yt-dlp/ffmpeg) is inconsistent and unexplained.
Instruction Scope
SKILL.md and the CLI scripts instruct the agent to import and run a VideoReaderMCP implementation located at workspace-code-dev/video-reader/src — which is not included in the bundle. The scripts insert a path resolved via parents[4] (author's /home/ykl/.openclaw path) into sys.path, meaning runtime will load and execute external code from the host filesystem. The instructions also assume use of yt-dlp/ffmpeg and Whisper; they reference .env and other local paths. Directing the agent to import and execute code outside the shipped files and to rely on host paths is scope creep and a significant execution surface.
Install Mechanism
There is no install spec (instruction-only plus two small scripts), which is low-risk generally. However, because the package lacks the actual 'src' implementation and relies on external tools and models (openai-whisper download), the runtime behavior depends on host environment installs and network downloads (models), which should be considered when evaluating runtime risk.
Credentials
SKILL.md documents required env vars (MINIMAX_API_KEY required, WHISPER_API_KEY when using API mode, plus optional LLM_BASE_URL/LLM_MODEL) but the registry lists none. Requiring bearer tokens for LLMs/Whisper is expected for this capability, but omitting them from the declared requirements is an incoherence that can hide credential needs. Also the skill can download models (~139MB) in local Whisper mode — a non-obvious resource/permission implication.
Persistence & Privilege
The skill is not marked always:true and does not request special platform privileges, which is OK. However, the scripts' hardcoded path arithmetic (parents[4] → /home/ykl/.openclaw/) and adding workspace-code-dev/video-reader/src to PYTHONPATH means the skill intentionally loads code from a specific location on the host. That gives it the ability to execute arbitrary code present at that path and to access the user's workspace files (temp/, .env, etc.), which raises privilege and persistence concerns relative to the stated purpose.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install video-content-watcher - 安装完成后,直接呼叫该 Skill 的名称或使用
/video-content-watcher触发 - 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
v1.0.0
Initial release of video-content-watcher.
- Extracts, transcribes, and analyzes video content from YouTube, Bilibili, and local files.
- Generates structured Markdown analysis reports via MiniMax LLM.
- Supports Chinese and English video analysis.
- Includes CLI and Python API for processing videos or raw text.
- Flexible configuration with environment variables for model and transcription settings.
元数据
常见问题
video-content-watcher 是什么?
Extract core content and generate structured analysis reports from YouTube, Bilibili, or local video files. Triggers when user asks to summarize, analyze, ex... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 124 次。
如何安装 video-content-watcher?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install video-content-watcher」即可一键安装,无需额外配置。
video-content-watcher 是免费的吗?
是的,video-content-watcher 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
video-content-watcher 支持哪些平台?
video-content-watcher 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-content-watcher?
由 Franklinyung(@franklinyung)开发并维护,当前版本 v1.0.0。
推荐 Skills