ASR

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

• Third-party content exposure detected (high risk: 0.80). The skill ingests and transcribes arbitrary user-provided audio (e.g., files read from directories and uploads handled by the Express /api/transcribe endpoint and fs.readFileSync), which is untrusted user-generated content that the agent will read and interpret as part of its workflow.