agent-immune

AI agent security: prompt injection detection, semantic memory, output scanning, prompt hardening