短视频黄金 3 秒钩子生成器

智能生成适用于抖音快手等短视频平台的主题钩子，支持分类筛选和数据追踪，助力高完播高流量内容创作。

This package appears internally consistent with a local short-video hook generator. Before installing, consider: 1) Verify provenance — the package lists no homepage and the source is 'unknown'; review the repository/author if possible. 2) Data stored locally — usage.json and usage_log.json (under the skill/data folder) record topics and (optionally) user_id; if you care about privacy, inspect/clear these files or modify monitor.py to avoid logging user_id. 3) There is no network exfiltration in the included code, but if you or the platform later wire the skill to analytics/payment backends (DEPLOY/PUBLISH docs recommend configuring payments/analytics), review those integrations carefully. 4) The SKILL.md claim of 'no sensitive words / platform-safe' is a policy statement, not enforced by code — do not rely on it for compliance. 5) For safety, run the skill in a sandbox or local environment first (python3 is invoked via index.js; on Windows 'python3' may not exist). If you want higher assurance, ask the developer for an upstream repository link or a signed release and confirm there's no hidden network behavior.

Type: OpenClaw Skill Name: shortvideo-hook Version: 1.1.1 The OpenClaw AgentSkills bundle 'shortvideo-hook' is classified as benign. The skill's core functionality involves generating short video hooks, utilizing local file I/O (e.g., `data/usage.json`, `data/usage_log.json`) for usage tracking and anonymous product analytics, as seen in `main.py` and `monitor.py`. The `index.js` wrapper safely executes the Python script by passing arguments as a JSON string, mitigating direct shell injection risks. All documentation files (`SKILL.md`, `README.md`, `BUSINESS.md`, etc.) are purely descriptive or instructional for human developers/users and contain no evidence of prompt injection attempts against the AI agent, nor any malicious instructions, data exfiltration, or persistence mechanisms.