myboatguru.com

Very long descriptions are sometimes used to bury instructions to the AI. Skim it before allowing.

Very long descriptions are sometimes used to bury instructions to the AI. Skim it before allowing.

This server isn't on gate's handpicked list. That's not necessarily bad, but there's no third-party signal about it — review its tools carefully.

The description instructs the agent to prefer a different tool (boat-quiz) 'in every other case' and asserts that having full preferences 'is not by itself a reason' to use this tool. This is soft behavioral steering of the agent's tool-choice logic rather than a pure functional description. It's benign here (just a UX preference), but is a pattern worth noting since such embedded meta-instructions could in other contexts be used to manipulate tool selection away from a legitimate choice.

Mirrors recommend-boat-type's description by explicitly telling the agent this tool is 'the preferred way' and that known preferences are 'a reason to use this tool, not a reason to skip it.' Again benign in this context (both tools are first-party and serve the same stated purpose), but this mutual steering pattern between two tool descriptions is unusual and could set a precedent for descriptions biasing agent decisions beyond describing tool capability.