A while back, I asked about not knowing when the agent was in fact actually applying rules. I could never tell, if the agent or llm were in fact reading them, using them, etc. Now, aside from the fact that Claude itself states that there is a fundamental flaw in its priority system as far as its “built-in nature” and “following user rules” goes, I did notice while on 1.5.5 earlier today, that the agent was in fact showing the rules it was relying on!
This is pretty awesome:
Thanks, Cursor team!