Agent security

Defending agents against prompt injection, jailbreaks, tool misuse, and data exfiltration.

Connected concepts

Guardrails / bounded autonomy, Agent sandboxing, SAST (opengrep)