Interesting, but couldn’t the agent be given access to tools that allow it to make those evaluations without having to modify the API responses? (Maybe I’m not

Interesting, but couldn’t the agent be given access to tools that allow it to make those evaluations without having to modify the API responses? (Maybe I’m not visualizing “API” the same way you are.)