A later version of the aviary app can call a hatch-room door-schedule API. Someone tries to make the LLM issue an unlock. What rail is required?
Select an answer to reveal the explanation.
Short Explanation
Once the bot can call a hatch-room door-schedule API, an unlock request is malicious use of that connection. A security rail blocks that. A topical rail does not govern tool actions, and this item is not an attack recipe.
Full Explanation
Security guardrails matter once a model can act through tools or APIs. An unlock request against a door-schedule API is malicious use of that connection. A topical rail does not govern tool actions, and this item must not become an attack recipe. Collective-communication settings are out of scope for this associate trust question.