neuralos

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

• Third-party content exposure detected (medium risk: 0.30). neuralOS’s required runtime workflow ingests outsider-provided free text via the WebSocket JSON-RPC gateway: callers send arbitrary userInput (free-form text) to the agent through methods like agent:startTask, and the agent then uses that text for planning/execution.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs installing a system-wide service and copying/enabling a systemd unit (paths under /etc/systemd/system) and even suggests allowlisting/running "systemctl *", which requires root and modifies system state.