Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations declare the operation non-read-only, non-destructive, and closed-world, so the description's cost note ('does not spend image tokens') and the 1-50 batch bound add real context. It never says whether prompts are persisted, what 'sync' or 'analyze' write, or what any call returns, which matters for a flagged non-read-only tool.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.