crisis-and-moderation

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

• Third-party content exposure detected (high risk: 0.70). The required workflow ingests outsider-authored free text in the form of “what actually happened from the user” / crisis comments (e.g., complaints, misinformation, deepfakes) that the agent must triage and draft responses to, so that user-provided narrative can be included in the LLM context and may contain indirect prompt-injection.