cloudflare-agentic-inbox

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

• Third-party content exposure detected (high risk: 1.00). The skill ingests untrusted, user-generated emails via the Email Routing handler (worker/index.ts -> message.text() stored to the mailbox) which the EmailAgentDurableObject exposes to the agent through the read_inbox tool (and the agent can act via send_email), so external email content can be read and influence agent actions.