sales-devi

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

• Third-party content exposure detected (high risk: 0.95). Outsider free text can enter the LLM context via Devi’s UI “Intent Detection” and “OutreachAI” modules, which read monitored posts/comments from external platforms (e.g., Facebook/Reddit/LinkedIn/X) and then use an LLM to classify intent and draft replies; those posts are authored by non-operating users.