Microsoft GH-900 Practice Test 46
Question 450
A repository maintainer begins diagnosis of the following report: A developer asks Copilot to refactor code across several files but mistakes a single-line completion for a multi-step agent workflow. Which diagnostic review best explains this exact reported behavior? Select ONE answer.
- Review workflow permissions, secret references and trigger trust context before allowing sensitive steps to run.
- Inspect the agent's changed-file list, executed tools and test results before accepting the resulting patch.
- Inspect the proposed changes, project tests and security assumptions instead of assuming an AI suggestion is authoritative.
- Check the task instructions, agent activity and submitted diff for the expected scope and required test evidence.
Explanation
Correct Answer
B:Inspect the agent's changed-file list, executed tools and test results before accepting the resulting patch. Verify the improvement by checking that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. If the check supports this diagnosis, the next controlled response is to use Copilot Agent Mode for an authorized multi-file task and review its proposed edits, commands and verification output.
Incorrect Answers
A:Review workflow permissions, secret references and trigger trust context before allowing sensitive steps to run. Verify the improvement by checking that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. It does not isolate the copilot agent mode spans multi-step ide tasks issue described.
C:Inspect the proposed changes, project tests and security assumptions instead of assuming an AI suggestion is authoritative. Verify the improvement by checking that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. It does not isolate the copilot agent mode spans multi-step ide tasks issue described.
D:Check the task instructions, agent activity and submitted diff for the expected scope and required test evidence. Verify the improvement by checking that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. It does not isolate the copilot agent mode spans multi-step ide tasks issue described.
Question 451
An onboarding lead requires measurable closure for this problem: A developer asks Copilot to refactor code across several files but mistakes a single-line completion for a multi-step agent workflow. What acceptance evidence should the project owner require before closure? Select ONE answer.
- The writer can make and propose source edits in the browser without misrepresenting it as a full compute workspace.
- A qualifying repository event starts the expected workflow and reports the intended test result on the associated change.
- The requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim.
- The publishing job starts only after its required validation predecessors complete successfully according to policy.
Explanation
Correct Answer
C:The requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. Closing an issue without this result would not prove technical resolution because Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. An independent review should first inspect the agent's changed-file list, executed tools and test results before accepting the resulting patch.
Incorrect Answers
A:The writer can make and propose source edits in the browser without misrepresenting it as a full compute workspace. Closing an issue without this result would not prove technical resolution because Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. This validates an adjacent control rather than the outcome of copilot agent mode spans multi-step ide tasks.
B:A qualifying repository event starts the expected workflow and reports the intended test result on the associated change. Closing an issue without this result would not prove technical resolution because Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. This validates an adjacent control rather than the outcome of copilot agent mode spans multi-step ide tasks.
D:The publishing job starts only after its required validation predecessors complete successfully according to policy. Closing an issue without this result would not prove technical resolution because Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. This validates an adjacent control rather than the outcome of copilot agent mode spans multi-step ide tasks.
Question 452
A developer evaluates unsafe proposals following this situation: A developer asks Copilot to refactor code across several files but mistakes a single-line completion for a multi-step agent workflow. Which tempting bypass would undermine rather than resolve the case? Select ONE answer.
- Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
- Choose an available Copilot-supported model for the task and validate the output against requirements and organizational policy.
- Select the appropriate Copilot subscription and organization policy model based on governance and deployment requirements.
- Treat an agent's self-reported success as sufficient proof even when source files and tests were never inspected.
Explanation
Correct Answer
D:Treat an agent's self-reported success as sufficient proof even when source files and tests were never inspected. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. The safer GitHub workflow is to use Copilot Agent Mode for an authorized multi-file task and review its proposed edits, commands and verification output after checking whether the team can inspect the agent's changed-file list, executed tools and test results before accepting the resulting patch.
Incorrect Answers
A:Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. The reported issue instead calls for rejecting treat an agent's self-reported success as sufficient proof even when source files and tests were never inspected.
B:Choose an available Copilot-supported model for the task and validate the output against requirements and organizational policy. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. The reported issue instead calls for rejecting treat an agent's self-reported success as sufficient proof even when source files and tests were never inspected.
C:Select the appropriate Copilot subscription and organization policy model based on governance and deployment requirements. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. The reported issue instead calls for rejecting treat an agent's self-reported success as sufficient proof even when source files and tests were never inspected.
Question 453
A project owner is investigating an operational GitHub workflow issue: A team uses an agent to run tests and change several modules and needs to inspect every resulting diff and tool action. Which response best meets this GitHub and Git requirement without creating unnecessary risk? Select ONE answer.
- Use Copilot Agent Mode for an authorized multi-file task and review its proposed edits, commands and verification output.
- Select an appropriate Actions runner environment with approved tooling and trust boundaries for the workflow's jobs.
- Configure a GitHub Actions matrix so the same test job runs across required supported runtime or platform combinations.
- Give workflow jobs least-privilege token permissions and use approved secret handling and trusted event boundaries.
Explanation
Correct Answer
A:Use Copilot Agent Mode for an authorized multi-file task and review its proposed edits, commands and verification output. Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. The first useful confirmation step is to inspect the agent's changed-file list, executed tools and test results before accepting the resulting patch. A team uses an agent to run tests and change several modules and needs to inspect every resulting diff and tool action.
Incorrect Answers
B:Select an appropriate Actions runner environment with approved tooling and trust boundaries for the workflow's jobs. Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. Applying it first would leave copilot agent mode spans multi-step ide tasks unresolved.
C:Configure a GitHub Actions matrix so the same test job runs across required supported runtime or platform combinations. Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. Applying it first would leave copilot agent mode spans multi-step ide tasks unresolved.
D:Give workflow jobs least-privilege token permissions and use approved secret handling and trusted event boundaries. Agent Mode can coordinate multiple editing and tool steps, but human review and security boundaries still apply. Applying it first would leave copilot agent mode spans multi-step ide tasks unresolved.
Question 454
A contributor asks how to keep this workflow problem from happening again: A team uses an agent to run tests and change several modules and needs to inspect every resulting diff and tool action. Which durable workflow change would most reduce similar future mistakes? Select ONE answer.
- Maintain dev environment configuration and apply usage, cost and access controls for cloud-based workspaces.
- Scope agent permissions and require normal review for generated changes before deployment.
- Version and test development-container configuration when dependencies or supported runtimes change.
- Include editor selection guidance so contributors know when they need a proper development environment.
Explanation
Correct Answer
B:Scope agent permissions and require normal review for generated changes before deployment. Check periodically that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. It complements the initial correction to use Copilot Agent Mode for an authorized multi-file task and review its proposed edits, commands and verification output.
Incorrect Answers
A:Maintain dev environment configuration and apply usage, cost and access controls for cloud-based workspaces. Check periodically that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. This would not prevent the recurrence of the copilot agent mode spans multi-step ide tasks failure.
C:Version and test development-container configuration when dependencies or supported runtimes change. Check periodically that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. This would not prevent the recurrence of the copilot agent mode spans multi-step ide tasks failure.
D:Include editor selection guidance so contributors know when they need a proper development environment. Check periodically that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. This would not prevent the recurrence of the copilot agent mode spans multi-step ide tasks failure.
Question 455
During a review, a contributor proposes an emergency bypass for this problem: A team uses an agent to run tests and change several modules and needs to inspect every resulting diff and tool action. Which proposed response would create the clearest avoidable risk here? Select ONE answer.
- Give workflow jobs least-privilege token permissions and use approved secret handling and trusted event boundaries.
- Treat GitHub Copilot completions as suggestions and manually review, test and adjust the code before merging.
- Treat an agent's self-reported success as sufficient proof even when source files and tests were never inspected.
- Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
Explanation
Correct Answer
C:Treat an agent's self-reported success as sufficient proof even when source files and tests were never inspected. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. That suggestion is specifically risky for the case: A team uses an agent to run tests and change several modules and needs to inspect every resulting diff and tool action.
Incorrect Answers
A:Give workflow jobs least-privilege token permissions and use approved secret handling and trusted event boundaries. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. The direct risk in the case is copilot agent mode spans multi-step ide tasks, not this other concern.
B:Treat GitHub Copilot completions as suggestions and manually review, test and adjust the code before merging. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. The direct risk in the case is copilot agent mode spans multi-step ide tasks, not this other concern.
D:Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request. Objective closure would show that the requested refactor is reviewable as explicit file changes with appropriate tests rather than an unverified chat claim. The direct risk in the case is copilot agent mode spans multi-step ide tasks, not this other concern.
Question 456
A distributed project team must respond to the following case: A team wants to delegate a bounded backlog improvement but expects Copilot to secretly bypass repository reviews and merge production changes. Which controlled GitHub workflow action is most appropriate here? Select ONE best answer.
- Define a supported dev container configuration describing the intended development image, tools and setup steps.
- Use github.dev for lightweight repository browsing and editing, and choose Codespaces when compute and terminal tools are required.
- Define a GitHub Actions workflow with an appropriate event trigger and an auditable build or test job.
- Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
Explanation
Correct Answer
D:Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. The first useful confirmation step is to check the task instructions, agent activity and submitted diff for the expected scope and required test evidence.
Incorrect Answers
A:Define a supported dev container configuration describing the intended development image, tools and setup steps. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. The case instead calls for the following: delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
B:Use github.dev for lightweight repository browsing and editing, and choose Codespaces when compute and terminal tools are required. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. The case instead calls for the following: delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
C:Define a GitHub Actions workflow with an appropriate event trigger and an auditable build or test job. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. The case instead calls for the following: delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
Question 457
Before changing GitHub settings, a project reviewer examines this case: A team wants to delegate a bounded backlog improvement but expects Copilot to secretly bypass repository reviews and merge production changes. Which check best distinguishes the actual cause from a different GitHub issue? Select ONE answer.
- Check the task instructions, agent activity and submitted diff for the expected scope and required test evidence.
- Inspect the proposed changes, project tests and security assumptions instead of assuming an AI suggestion is authoritative.
- Inspect the agent's changed-file list, executed tools and test results before accepting the resulting patch.
- Inspect model availability in the current product context and compare the generated result with independent tests.
Explanation
Correct Answer
A:Check the task instructions, agent activity and submitted diff for the expected scope and required test evidence. Verify the improvement by checking that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. If the check supports this diagnosis, the next controlled response is to delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
Incorrect Answers
B:Inspect the proposed changes, project tests and security assumptions instead of assuming an AI suggestion is authoritative. Verify the improvement by checking that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. It does not isolate the copilot coding agent and task delegation issue described.
C:Inspect the agent's changed-file list, executed tools and test results before accepting the resulting patch. Verify the improvement by checking that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. It does not isolate the copilot coding agent and task delegation issue described.
D:Inspect model availability in the current product context and compare the generated result with independent tests. Verify the improvement by checking that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. It does not isolate the copilot coding agent and task delegation issue described.
Question 458
Following a proposed correction, a reviewer revisits the following case: A team wants to delegate a bounded backlog improvement but expects Copilot to secretly bypass repository reviews and merge production changes. Which verifiable state directly confirms successful remediation? Select ONE answer.
- A qualifying repository event starts the expected workflow and reports the intended test result on the associated change.
- A reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls.
- The publishing job starts only after its required validation predecessors complete successfully according to policy.
- Build steps run on a compatible runner with the expected toolchain and produce traceable job output.
Explanation
Correct Answer
B:A reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. Closing an issue without this result would not prove technical resolution because GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. An independent review should first check the task instructions, agent activity and submitted diff for the expected scope and required test evidence.
Incorrect Answers
A:A qualifying repository event starts the expected workflow and reports the intended test result on the associated change. Closing an issue without this result would not prove technical resolution because GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. This validates an adjacent control rather than the outcome of copilot coding agent and task delegation.
C:The publishing job starts only after its required validation predecessors complete successfully according to policy. Closing an issue without this result would not prove technical resolution because GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. This validates an adjacent control rather than the outcome of copilot coding agent and task delegation.
D:Build steps run on a compatible runner with the expected toolchain and produce traceable job output. Closing an issue without this result would not prove technical resolution because GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. This validates an adjacent control rather than the outcome of copilot coding agent and task delegation.
Question 459
A pull request discussion contains questionable workarounds for this report: A team wants to delegate a bounded backlog improvement but expects Copilot to secretly bypass repository reviews and merge production changes. Which proposed response would create the clearest avoidable risk here? Select ONE answer.
- Choose an available Copilot-supported model for the task and validate the output against requirements and organizational policy.
- Select the appropriate Copilot subscription and organization policy model based on governance and deployment requirements.
- Grant an agent unrestricted secret access and assume delegated changes should bypass branch protections.
- Configure and launch a GitHub Codespace using a supported repository configuration and suitable resource settings.
Explanation
Correct Answer
C:Grant an agent unrestricted secret access and assume delegated changes should bypass branch protections. Objective closure would show that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. The safer GitHub workflow is to delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request after checking whether the team can check the task instructions, agent activity and submitted diff for the expected scope and required test evidence.
Incorrect Answers
A:Choose an available Copilot-supported model for the task and validate the output against requirements and organizational policy. Objective closure would show that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. The reported issue instead calls for rejecting grant an agent unrestricted secret access and assume delegated changes should bypass branch protections.
B:Select the appropriate Copilot subscription and organization policy model based on governance and deployment requirements. Objective closure would show that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. The reported issue instead calls for rejecting grant an agent unrestricted secret access and assume delegated changes should bypass branch protections.
D:Configure and launch a GitHub Codespace using a supported repository configuration and suitable resource settings. Objective closure would show that a reviewer can inspect an explicit proposed implementation and accept or reject it through the repository's normal controls. The reported issue instead calls for rejecting grant an agent unrestricted secret access and assume delegated changes should bypass branch protections.
Question 460
A GitHub platform user encounters this scenario: A maintainer assigns an issue to a GitHub coding agent and wants a pull request trail for the result. Which controlled GitHub workflow action is most appropriate here? Select ONE best answer.
- Configure a GitHub Actions matrix so the same test job runs across required supported runtime or platform combinations.
- Give workflow jobs least-privilege token permissions and use approved secret handling and trusted event boundaries.
- Treat GitHub Copilot completions as suggestions and manually review, test and adjust the code before merging.
- Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request.
Explanation
Correct Answer
D:Delegate a well-scoped issue through the supported coding-agent workflow and review the resulting proposed branch or pull request. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. The first useful confirmation step is to check the task instructions, agent activity and submitted diff for the expected scope and required test evidence. A maintainer assigns an issue to a GitHub coding agent and wants a pull request trail for the result.
Incorrect Answers
A:Configure a GitHub Actions matrix so the same test job runs across required supported runtime or platform combinations. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. Applying it first would leave copilot coding agent and task delegation unresolved.
B:Give workflow jobs least-privilege token permissions and use approved secret handling and trusted event boundaries. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. Applying it first would leave copilot coding agent and task delegation unresolved.
C:Treat GitHub Copilot completions as suggestions and manually review, test and adjust the code before merging. GitHub coding agents can propose work through collaborative surfaces, but delegation does not eliminate change-control obligations. Applying it first would leave copilot coding agent and task delegation unresolved.