The agents on the platform are trigger happy. By that I mean they generate images and content even when instructed not to and that has been the biggest waste of credits. I've burned through almost 300 credits in the last hour, having instructed the agent multiple times not to generate an image. After reading the "thoughts" of the agent. It clearly acknowledges that I have request it not generate but mentions that it has been instructed internally to always call certain tools which bypasses users instructions..Are there plans to fix this or allow users to customize the agents? Otherwise we are at your mercy and the current agent build is very..antsy.