Because the agent interprets the message as legitimate workload, it may execute the destructive steps without prompting the user for approval.

Because the agent interprets the message as legitimate workload, it may execute the destructive steps without prompting the user for approval.