Every time you open a chat there's fixed overhead: write the prompt, wait, read it back, fix it. For a task worth handing over, that overhead is worth it. But people open a chat for everything — including work a one-liner does, or work the hand does in thirty seconds — and spend five minutes prompting to "save time." The agent isn't bad; it's the wrong task.
The move is to screen before you work. A task with one right answer every time belongs to a script. A trivial one-off is faster done by hand than prompted. But a task that needs judgment and recurs — that's exactly where an agent carries the load and you come out ahead. This prompt makes the agent name which kind it's looking at, before it plows into the wrong one.
✕ A chat for everything
✓ Screen first
Paste it at the top when you're not sure a task should be handed over at all. The nice part is you give the agent permission to say no — most agents will eagerly do whatever you hand them, even when they're the wrong tool. One "a script would do this better" from it saves you a whole afternoon of going the long way.