yeah prompts and skills and roles can encapsulate instruction modules or standard operating procedures that are not obvious in the weights themselves but if you’re naming your agents or modeling them as “adversaries” you’re mostly just playing house