Question 1
At Aster Retail, a storefront service desk is improving order-support before a seasonal service launch (SF-M03-01). The administrator observes that a service agent delegates a case to a specialized billing agent without the required case identifier. Any fix must ensure that customer order details stay within each authorized account. Which observable result would best demonstrate that the correction actually worked?
Explanation
Correct Answer
This is the direct acceptance signal for the reported failure: In a controlled Case acceptance test, the receiving specialist obtains only the intended case context and returns a traceable outcome; the record is retained under Aster Retail's rollout (SF-M03-01). The chosen correction was to define approved handoff inputs and evidence needed by the recipient agent; the measurement therefore tests the relevant source, routing or authorization outcome rather than unrelated activity.
Incorrect Answers
The proposed observation “The single-agent implementation answers supported FAQs without orchestration overhead; confirm this and compare with approved baseline behavior” would check a different original case: A basic FAQ workflow has one data source and no specialized actions yet architects propose six coordinating agents. The correction must ensure that supplier prices are never shown outside entitled teams. Its mechanism is: Multiple agents should be chosen for a business or control need, not novelty. In the present case, a service agent delegates a case to a specialized billing agent without the required case identifier. The correction must ensure that customer order details stay within each authorized account; the more direct acceptance evidence is “The receiving specialist obtains only the intended case context and returns a traceable outcome; confirm this while testing the unauthorized case” because multi-agent coordination depends on explicit context and controlled responsibility transfer.
The proposed observation “Tests show the service agent delegates only permitted finance subtasks without crossing access boundaries; confirm this using a least-privilege test identity” would check a different original case: A support and finance workflow has distinct permissions and responsibilities that one broad agent cannot govern safely. The correction must ensure that student records never cross program authorization boundaries. Its mechanism is: Multi-agent design can separate responsibilities but increases orchestration complexity. In the present case, a service agent delegates a case to a specialized billing agent without the required case identifier. The correction must ensure that customer order details stay within each authorized account; the more direct acceptance evidence is “The receiving specialist obtains only the intended case context and returns a traceable outcome; confirm this while testing the unauthorized case” because multi-agent coordination depends on explicit context and controlled responsibility transfer.
The proposed observation “Attempts to invoke forbidden tools fail and allowed calls produce audit records; confirm this and retain a reproducible test record” would check a different original case: An Agentforce agent is granted a generic MCP tool endpoint that can mutate unrelated customer systems. The correction must ensure that escalations preserve only the needed guest context. Its mechanism is: Open tool protocols do not remove the need for least-privilege access boundaries. In the present case, a service agent delegates a case to a specialized billing agent without the required case identifier. The correction must ensure that customer order details stay within each authorized account; the more direct acceptance evidence is “The receiving specialist obtains only the intended case context and returns a traceable outcome; confirm this while testing the unauthorized case” because multi-agent coordination depends on explicit context and controlled responsibility transfer.