Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward YouTube clip-planning prompt with no bundled code, persistence, credential use, or hidden behavior.
Installers should treat this as a content-analysis helper: only provide media or transcripts you have rights to process, avoid confidential/private recordings unless you are comfortable sharing them with the active agent/tools, and specify the desired language or locale if you do not want Japanese-style clip titles or tags.
The skill solicits YouTube URLs, MP4 paths, and transcripts, and optionally produces Whisper-derived subtitles, but it does not warn users about privacy, copyright, consent, or local-file disclosure risks. This can lead users to provide sensitive or unauthorized media, creating compliance and privacy exposure even though the skill itself is not overtly exfiltrating data.
The example output and scoring/composition rules strongly bias titles, tags, and emphasis terms toward Japanese text and Japanese clip-culture conventions, but the document never says this is optional or limited to a Japan-specific workflow. That can violate language/locale policy because it effectively forces a specific language style without user opt-in.
No suspicious patterns detected.
