ai-music

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill runtime ingests outsider-authored free text via the user’s --input JSON fields (prompt, tags, lyrics) and then forwards them to runcomfy run ... for model generation without requiring a prior selection of a specific outsider-authored item.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.70). The skill passes runtime "audio" URLs into inpaint/outpaint model calls (e.g., https://your-cdn.example/song.mp3 and https://your-cdn.example/hook-30s.mp3), which the CLI/Model API will fetch during execution and—per the skill's own warning—could contain embedded/steganographic content that influences generation (indirect prompt injection).