Comment by gas9S9zw3P9c

It depends. If you have an LLM that uses reasoning the explanation for why decisions are made can often be found in the reasoning token output. So if the agent later has access to that context it could see why a decision was made.