agentforce-observe

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

• Third-party content exposure detected (high risk: 0.95). The skill’s runtime LLM context is fed by STDM-derived conversation content (including outsider-authored user messages and agent outputs) via AgentforceOptimizeService.getConversationDetails() / getMultipleConversationDetails() (and optionally getLlmStepDetails()), which returns readable text/prompt/llm_response fields that the agent then summarizes/classifies—i.e., outsider free text from prior conversations enters the agent’s LLM context.