I built a system to catch unreliable AI agents. In my own evaluation, it missed the worst one.

My framework exists to identify unreliable transmitters in a multi-agent pipeline. In my own...