hermes-katana

Hermes Katana is a defense-in-depth security tool designed for AI agents, providing mechanisms for tracking input provenance, scanning content for prompt injections, and enforcing YAML policies before tool execution. Notable features include configurable human-in-the-loop escalation, purpose-trained injection classifiers, and a tamper-evident audit trail for decision-making, all aimed at enhancing the security posture of AI applications. This tool is particularly useful for developers looking to safeguard AI systems from potential vulnerabilities and malicious inputs.