infinitetalk

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

• Third-party content exposure detected (medium risk: 0.65). In the required runtime workflow, if --text is provided by an outsider, it flows into the TTS pipeline as free-form text via scripts/infer_infinitetalk.py:process_tts_single() → pipeline(text, ...), which then feeds into the model/LLM-like text-generation context (and the script logs a preview of it).