Developing in Agentic AI Systems v1.0

Case study -
This is a case study. Case studies are not timed separately. You can use as much exam time as you would like to complete each case. However, there may be additional case studies and sections on this exam. You must manage your time to ensure that you are able to complete all questions included on this exam in the time provided.
To answer the questions included in a case study, you will need to reference information that is provided in the case study. Case studies might contain exhibits and other resources that provide more information about the scenario that is described in the case study. Each question is independent of the other questions in this case study.
At the end of this case study, a review screen will appear. This screen allows you to review your answers and to make changes before you move to the next section of the exam. After you begin a new section, you cannot return to this section.

To start the case study -
To display the first question in this case study, click the Next button. Use the buttons in the left pane to explore the content of the case study before you answer the questions. Clicking these buttons displays information such as business requirements, existing environment, and problem statements. If the case study has an All Information tab, note that the information displayed is identical to the information displayed on the subsequent tabs. When you are ready to answer a question, click the Question button to return to the question.

Existing Environment -

GitHub Environment -
The GitHub environment contains the following:
Three repositories named product-api, billing-service, and infra-terraform.
Branch protection on the main branch in all repositories that requires at least one pull request review before merging
GitHub Actions runners used across all workflows
A GitHub team named SG_Dev that contains developers
A GitHub team named SG_Review that contains senior engineers and a security team
A .github/copilot-instructions.md file that includes general coding conventions for all features

Agent environment -
The product-api repository uses a GitHub Copilot coding agent named agent1 that has the following configurations:
No custom agent profile is defined.
A Model Context Protocol (MCP) server named MCP1 is deployed to https://mcp.litwareinc.internal and provides access to internal ticketing and deployment APIs. MCP1 requires an API key for authentication.
A second Copilot coding agent named agent2 handles changes in infra-terraform and runs in parallel with agent1 when both agents have open assigned issues.
Copilot memory is NOT enabled for the organization.

Problem Statements -
Litware identifies the following issues:
During two recent sessions, agent1 accessed files in billing-service, which is outside the agent's intended scope. agent1 makes code changes immediately after receiving a task.
A developer named Ben, who is on the SG_Dev team, reports that agent1 completed a session with a successful status and opened a pull request, but the pull request contains no file changes. Other developers report this intermittently as well.
Both agent1 and agent2 modified shared/config.yaml in a parallel test run, generating conflicting outputs. agent1 consistently uses raw try-catch blocks for error handling, which violates the defined implementation guidelines of SG_Dev.

Requirements -

Planned Changes -
Litware plans to make the following changes:
Ensure that agent1 can access all the tools in the environment.
Provide product-api with specific instructions to agent1 without affecting Copilot Chat or Copilot code review.
Configure MCP1 as a tool for agent1 by modifying the product-api repository MCP configuration.
Ensure that Copilot retains details that it has learned and uses that knowledge for future work. This must be applied to all licensed members of the organization.

Implementation guidelines -
The development team at Litware identifies the following implementation guidelines:
Agent workflows must be able to run in parallel.
Application error handling must use the repository ErrorHandler class. agent1 and agent2 must run on isolated branches during parallel execution. File-level conflicts must be detected before merges, and both agents must be able to run concurrently.

Security requirements -
Litware identifies the following security requirements:
Only the members of SG_Review must be able to approve agent1 plan outputs.
All API keys must be stored and accessed securely.
The developers must NOT be able to self-approve.

Agent configuration -

You need to troubleshoot the issue reported by Ben.
What should you review?

HOTSPOT -

Case study -
This is a case study. Case studies are not timed separately. You can use as much exam time as you would like to complete each case. However, there may be additional case studies and sections on this exam. You must manage your time to ensure that you are able to complete all questions included on this exam in the time provided.
To answer the questions included in a case study, you will need to reference information that is provided in the case study. Case studies might contain exhibits and other resources that provide more information about the scenario that is described in the case study. Each question is independent of the other questions in this case study.
At the end of this case study, a review screen will appear. This screen allows you to review your answers and to make changes before you move to the next section of the exam. After you begin a new section, you cannot return to this section.

To start the case study -
To display the first question in this case study, click the Next button. Use the buttons in the left pane to explore the content of the case study before you answer the questions. Clicking these buttons displays information such as business requirements, existing environment, and problem statements. If the case study has an All Information tab, note that the information displayed is identical to the information displayed on the subsequent tabs. When you are ready to answer a question, click the Question button to return to the question.

Overview -
Contoso Ltd. is a software development company located in the United States.

Existing Environment -

GitHub Environment -
Contoso uses GitHub Enterprise and assigns GitHub Copilot Pro+ licenses to its developers. The developers use Microsoft Visual Studio Code as their IDE.
Contoso has a customer portal. The code for the portal is stored in a GitHub repository named repo1 that contains the following:
A custom agent named agent1 that includes instructions to review specs related to best practices
A custom instruction file named validate-instructions.md that is used to validate tone of voice and applies to all .md and .txt files
A custom instruction file named codereview.instructions.md that is used by the Copilot coding agent but is excluded for use by the Copilot code review repo1 has the following structure:
The front-end is stored in the /frontend folder.
The API logic is stored in the /api folder.
Contoso has a second repository named repo2 that contains a legacy .NET application named App1 built by using .NET 6. repo2 has a multi-agent workflow for modernization tasks.
Contoso enables the Model Context Protocol (MCP) registry and allows the Microsoft Learn MCP Server. Every developer must configure their own connection to the Learn MCP Server.

Problem Statements -
The developers working in repo1 report that the Microsoft Learn documentation is NOT being retrieved when they attempt to validate a design by using agent1.
The testing team at Contoso identifies that the customer portal uses inconsistent UI styles, which leads to customer confusion and branding issues. The UI inconsistencies stem from variations in the folder structure.

Agent Logs -
You have the following logs for the multi-agent workflow used in repo2.


Requirements -

Planned Changes -
Contoso plans to have all agents and developers in repo1 use the Microsoft Learn MCP to ensure that reviews are validated by using the appropriate documentation. This must be implemented centrally.
Contoso plans to leverage AI-powered coding agents to implement new portal features and pages.

Technical Requirements -
App1 must be upgraded to .NET 10. A previous upgrade attempt was started by using the Copilot modernization agent, but the attempt was never finalized.
You plan to retry the upgrade. You must first analyze App1 by using AI, and then generate a report that contains breaking changes and deprecated patterns before retrying the upgrade.
All AI-generated code for UI styling must adhere to a predefined folder structure.
The architects at Contoso need help building implementation plans for repo1. The company wants to implement a new agent named agent2 to analyze the code base and the code requirements, and then respond with a detailed plan. The agent must NOT be able to edit files or run local commands.
The developers must be able to delegate work to the Copilot coding agent by assigning issues to the agent.
You are evaluating the logs of the multi-agent workflow in repo2.
For each of the following statements, select Yes if the statement is true. Otherwise, select No.
NOTE: Each correct selection is worth one point.

HOTSPOT -
You have the following agent logs.

For each of the following statements, select Yes if the statement is true. Otherwise, select No.
NOTE: Each correct selection is worth one point.

You have a GitHub repository that uses GitHub Copilot code review on pull requests.
You plan to add repository-wide code review guidance that will apply to all files.
You need Copilot code review to consistently apply the guidance during pull request reviews.
What should you do?

You have a GitHub Enterprise repository that uses the GitHub Copilot coding agent and opens draft pull requests for assigned issues.
Evaluation results show that the agent repeatedly opens pull requests that modify .github/workflows/*.yml to bypass failing checks instead of fixing the underlying code.
You need to ensure that the agent fixes the underlying code instead of bypassing the failing checks.
What should you do?

You use a GitHub Actions workflow to orchestrate a multi-agent draft-review process for pull requests. The workflow uploads a single combined review-package artifact at the end of each run.
You discover that the review-package artifact is occasionally missing from the workflow run history, and you suspect that a user deleted it.
You need to identify which user manually deleted a workflow run artifact and when the deletion occurred.
What should you use?

You have a GitHub Enterprise Cloud Organization that uses the GitHub Copilot coding agent.
Copilot creates a draft pull request for an assigned issue, and the pull request timeline shows Copilot started work.
After 70 minutes, the agent session log stops updating, and the pull request body status stops changing.
You need to restart the agent so that it continues the task from the issue context and produces new commits to the existing draft pull request.
What should you do?

You have a GitHub Enterprise Cloud Organization that uses a custom coding agent to run GitHub Actions workflows that create branches, open pull requests, and merge changes after required checks pass.
You need to log all agent-initiated actions and ensure that the logs are retained for two years.
What should you do?

You have multiple GitHub Copilot coding agents that run tasks concurrently.
You live stream the session log output and see the following.

What is a possible cause of the error?

You are analyzing the following agent logs.

You need to classify the error in a report for your company's CTO.
How should you classify the error?

You have multiple GitHub Copilot coding agents that run tasks concurrently.
You are monitoring the agents from the terminal by using the GitHub CLI.
An agent appears stalled.
You need to live stream the session log output.
What should you do?

You have a GitHub Copilot coding agent that has completed a pull request for a security fix in your repository.
Before merging, you need to evaluate the quality of the agent's work by using both automated evaluation signals and human review.
You review the session log and the pull request.
What are two automated evaluation signals generated by the coding agent's built-in scanning tools? Each correct answer presents a complete solution.
NOTE: Each correct selection is worth one point.

You have a GitHub Enterprise repository that runs an autonomous agent by using a GitHub Actions workflow. The workflow has the following jobs: agent-run that generates trace.json and plan.md review that waits for human approval before continuing deploy that uses the outputs from agent-run
You need to make the files inspectable in the GitHub Actions UI and ensure that the files are available to the review and deploy jobs.
What should you do in the workflow?

A developer uses the GitHub Copilot CLI in plan mode.
Copilot produces a plan.
What does Copilot do next?