Same rationale as the morning_briefing talent: these agents synthesize entity data into one-line context summaries, not deep relationship analysis. Brief mode cuts 95-99% of payload with no information loss for this use case.