newsletter-monitor

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

• Third-party content exposure detected (high risk: 0.85). This skill fetches messages from an AgentMail inbox at runtime and extracts their subject/body text into full_text and context_snippets, which are then emitted as JSON/summary (i.e., outsider-authored email free text flows into the agent’s context if the agent consumes this output for downstream steps).