What does it mean that agent scope should be narrow?

Each agent should handle one specific situation — multi-purpose agents are fragile.

Example: You create a 'health agent' that tries to handle sleep, exercise, hydration, posture, and meal timing. Within a week you stop following it because the trigger conditions overlap and contradict. When you're tired but haven't exercised, the agent gives you conflicting instructions. Replace it with five narrow agents — one per concern — and each fires cleanly because each has exactly one trigger and one action.

Try this: Pick one agent you currently run (or want to run) that handles more than one situation. Split it into two or three narrower agents, each with a single trigger condition and a single action. Write each one on a separate card or line. Test them independently for three days and notice which ones actually fire.