@dionysuz.eth
if an agent gets something wrong my favorite way to fix it rn is to ask it to figure out what in the context made it produce the wrong result, fix that, and feed the corrected context to another agent to actually fix it (or another iteration of fixing the context)