Salesforce Agentforce Specialist Practice Test

Answer choices

1. A. Assign the permission set to the deployment administrator and run the public agent under that administrator, for the specified implementation requirement.
2. B. Create the target-org agent user and assign the deployed permission set to that identity, under the organization’s defined implementation and exception-management process.
3. C. Add the permission names to the agent instructions so Atlas can apply them at runtime, for this requirement.
4. D. Publish the channel and rely on publication to create the missing identity and assignments, for the specified implementation requirement.

Wrong-answer review

• A. Assign the permission set to the deployment administrator and run the public agent under that administrator, for the specified implementation requirement.: Running a public agent as a broadly privileged deployment administrator violates the dedicated least-privileged identity design.
• C. Add the permission names to the agent instructions so Atlas can apply them at runtime, for this requirement.: Instructions cannot create platform users or perform permission-set assignments.
• D. Publish the channel and rely on publication to create the missing identity and assignments, for the specified implementation requirement.: Channel publication exposes the interaction surface but does not create an agent user or assign permissions.

Answer choices

1. A. Chunk the manual, organize the chunks in a search index, and use an authorized retriever to return relevant context, as selected.
2. B. Send the entire manual with every prompt because grounding requires the complete source document, as the proposed design for the complete governed operational workflow.
3. C. Store only the document title and ask the model to reconstruct the missing warranty language, within the described context.
4. D. Add the manual to the agent role because roles automatically perform indexed passage retrieval, under the stated technical, operational, and governance constraints.

Answer choices

1. A. Implement Data 360 with the structured and unstructured sources and configure advanced hybrid search, under the proposed approach.
2. B. Use a basic data library and periodically paste customer attributes into its uploaded documents, for the described technical objective.
3. C. Use merge fields alone because they perform semantic retrieval across product manuals, under the stated decision criteria.
4. D. Copy the data into prompt instructions so the model can perform real-time exact and semantic searches.

Answer choices

1. A. Enable and implement Data 360 for advanced grounding across the external unstructured contracts and structured customer data.
2. B. Upload monthly contract copies to a basic library and treat them as current zero-copy content, for the stated implementation and support requirements.
3. C. Place full contracts into prompt instructions and join them to customer data manually, under the described data 360 for agentforce criteria.
4. D. Use one merge field to retrieve and semantically search every external contract, as the selected response to the described condition.

Answer choices

1. A. Testing Center administration, because test execution grants prompt-template authoring rights, as the recommended response to this scenario.
2. B. Another execute-only prompt-template permission set, as the selected response to the described condition.
3. C. Agent-channel administration for the channels where the templates are used, under the organization’s defined implementation and exception-management process.
4. D. The Prompt Template Manager capabilities required to manage and execute prompt templates, within this design.

Answer choices

1. A. Session tracing for one successful conversation, under this approach.
2. B. Postdeployment adoption monitoring, as the selected approach for the stated technical and business outcome.
3. C. Testing Center with a repeatable suite of scenario-based cases, for consideration.
4. D. A production deployment followed by user feedback, within the stated policy framework.

Answer choices

1. A. Rely on harmful-output detection to remove any sensitive fields after the agent retrieves them.
2. B. Give both actions identical field access because the Trust Layer already monitors interactions, as selected.
3. C. Use Atlas routing to decide dynamically whether the requesting user deserves access, as the recommended response to this scenario.
4. D. Configure permission-based record and field access appropriate to each agent type and action, for the stated scenario.

Answer choices

1. A. A2A, because the carrier API should be treated as a specialized remote agent, within the described operational context.
2. B. MCP, because the assistant needs standardized access to an external tool or service, for the stated requirement.
3. C. Agent API, because every external API call must begin an Agentforce conversation session, when applied.
4. D. A new point-to-point connector for each assistant, avoiding a shared protocol, when applied.

Answer choices

1. A. They must share one retriever object because a search index can be referenced only once, within the documented operational, security, ownership, and validation requirements.
2. B. They have no retrievers because the model queries the common index directly, for the described technical objective and its associated operational control requirements.
3. C. Their prompts become identical because any shared index removes library-level configuration, as the organization’s selected response.
4. D. Each library retains its own unique retriever, even though both retrievers point to the same index, within cross-functional operational-accountability boundaries.

Answer choices

1. A. A reusable Prompt Builder Field Generation template grounded with the required Salesforce resources, within this design.
2. B. A separate Flex template for every Case so each record preserves its existing prompt text, under end-to-end security-and-governance requirements.
3. C. A static Knowledge article copied into the field without current Case grounding, for the required prompt engineering outcome.
4. D. An ungoverned free-form prompt that each representative rewrites before saving the field, for this requirement.