Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations only mark it non-read-only and idempotent; the description adds the concrete consequences: an auth scope requirement, that a chart may be saved, that an omitted start defaults to today in the birthplace timezone, that an empty preview can be returned explicitly, and that no model generation or payment is initiated. It also flags names as untrusted content, which is a genuine safety disclosure absent from the annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.