A Ridgeline custom agent action needs to answer a policy question by pulling from indexed clinic documents. How is the retriever most likely incorporated into this action's execution?
Select an answer to reveal the explanation.
Short Explanation
The action leans on a prompt template behind the scenes, and that template is where the retriever is wired in as a grounding source — so by the time the model responds, the relevant policy content is already sitting in its context.
Full Explanation
A custom agent action that needs to answer a policy question typically invokes a prompt template as its backing mechanism, and it's within that prompt template's configuration that a retriever is set as a grounding source — meaning retrieved, indexed content becomes available in context when the model generates its response for that action. Claiming a retriever can only be used by a human running a report ignores its actual purpose, which is real-time grounding for conversational agent responses, not a separate offline reporting tool. There's no mechanism where a retriever runs independently and emails results to an administrator — that describes a different kind of automation entirely, unrelated to how grounding is wired into agent actions. Bypassing retrieval and answering only from general LLM knowledge is exactly the ungrounded approach that risks inaccurate, non-org-specific answers, which is the problem grounding via retrievers is meant to solve for policy questions. For the exam, connect the dots: agent action → prompt template → configured retriever as grounding source → grounded response.