●builderYou can use these attack categories to harden your agent's tool-use and file-access capabilities.
●researcherThis establishes a rigorous testing ground for evaluating security vulnerabilities in agentic workflows.
●policyThis provides a standardized way to measure the safety and alignment of agents in business environments.