AI Tasks are sometimes very frustrating, more task examples would be better to train the Ai and a Rating of the execution would be great too. It would be direct feedback after the execution so that the Ai knows if there is room for improvment or not.

Just how would you implement this? Botpress is not training anything - it is all prompting and RAG. Training would require a custom model at a cost that you likely would get pale from. It would mean using at the end custom models. it is a good idea - but right now, especially wit hthe samll context version, you can not even inject a proper RAG base set of guidelines - not enough context space.