Security firewall for AI agents — block prompt injection, data exfiltration, and unauthorized tool access.
Scoring was not performed