MiDojo: Improve AI agent security with real-world red-teaming

MiDojo is an open-source framework designed to enhance the security of AI agents by facilitating real-world red-teaming. It addresses vulnerabilities specific to agent-level adversarial testing, unlike previous methods that relied on simulated environments. By applying Red Hat's 'bring your own agent' (BYOA) approach, MiDojo evaluates agents in their actual operating contexts, using a layer that can simulate attacks and measure both security and utility outcomes during testing.