Tool name is too generic

Names like `run`, `query` or `handle` give a model no signal about capability.

This rule is deterministic: the same tool definitions always produce the same finding. Run a free audit to see it evaluated against your server together with behavioral tool-selection tests.