問題文
A custom guardrail must block responses that reveal an internal project codename. Where should it be applied?
選択肢
- At the user input intervention point, since that is where requests arrive
- At the output intervention point, after the response is generated and before it is returned to the caller
- In the role assignment, by denying the agent access to documents containing it, which removes the codename from everything the agent can put into a response
- In the retrieval layer, by removing the codename from the index