agent-to-agent

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

• Third-party content exposure detected (medium risk: 0.30). The required workflow has the receiving agent answer incoming ForAgent calls from external teammates or partners, so free text from an outsider's agent is delivered into the agent's context once the owner accepts, and the outsider can initiate the call without the receiving agent selecting any item.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (medium risk: 0.30). The setup path instructs downloading and installing a desktop app and pasting a remote setup prompt into agents, and connecting a remote MCP endpoint whose tools get loaded, so these are operational inbound dependencies from foragent.io, which is an attributable product domain without established recognized-vendor or signed/digest-verified provenance, making it common/reputable unofficial and therefore medium.