speech

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime path ingests outsider-authored free text from --input-file / --input (or speak-batch JSONL lines via --input), and that text is sent into the OpenAI TTS request as the input field (and optionally instructions), so any text the operating user didn’t author can become LLM-context via the model call.