Which statement about how an Agentforce agent respects Salesforce security when it executes an action is true?
- AThe agent runs its actions as a defined agent user and is constrained by that user's profile and permission set access. Correct
- BThe agent bypasses object and field-level security because it is an internal platform process rather than a user.
- CThe agent inherits the profile of whichever end user is chatting, changing its access on every conversation.
- DThe agent enforces organisation-wide defaults only and disregards field-level security entirely when it reads data.
Why A is correct: Correct. Agent actions execute in the context of an assigned user, so object, field, and record access follow that user's profile and permission sets, and the agent cannot surface data the user could not see.
Why B is wrong: Agents do not bypass the security model. This tempts candidates who assume automation ignores permissions, but agent actions are bound to a user's access.
Why C is wrong: The agent operates under its own configured agent user, not the chatting user's profile. The option is plausible because access does matter, but the running context is fixed to the agent user.
Why D is wrong: Field-level security still applies to agent actions, so ignoring it is incorrect. Naming org-wide defaults makes this sound informed while omitting the layers that also apply.