Open-source adversary emulation for AI agents and MCP servers.

Enables agents and applications to detect, inspect, and neutralize prompt injection, jailbreak attempts, and hidden instruction escapes in untrusted text before it is passed into prompts.