That paper is hard to evaluate because their modal example of a “context file” is a bad practice that arose from early attempts at text-based agent guidance bef

That paper is hard to evaluate because their modal example of a “context file” is a bad practice that arose from early attempts at text-based agent guidance before we recognized that contextual instruction was the aim, not just “context.” With a “context file” you’re almost guaranteed to add bloat without useful behavior change because it’s just a pre-set list of things that could be good to know about. So the results of the study don’t generalize to every text file used for instruction or even most.