We built a runtime security layer for AI agents (instead of prompt filtering)(github.com)1pointsbyArikernel2 hours ago1comments