Question 170
The current design is: Use a bounded multi-file editing workflow and manually review all proposed edits rather than authorizing agent tool execution. The prior failure was: An unconstrained agent attempts unrelated shell operations during a simple text refactor. Which ONE ongoing control best protects against recurrence of that specific failure?
- Keep branch protections and required checks for agent-created pull requests.
- Review agent tool approvals, modified files and generated tests before merging.
- Choose the least-privileged Copilot interaction mode for the requested work.
- Keep custom agent scopes explicit and check output against human review requirements.
Explanation
Correct Answer
C:The selected workflow granted broader tools than the task required. Choose the least-privileged Copilot interaction mode for the requested work.
Incorrect Answers
A:Keep branch protections and required checks for agent-created pull requests. It would leave tested AI-assisted changes that preserve stated requirements without the specified safeguard.
B:Generated code requires independently defined acceptance tests and review against the real program contract. Preventing this failure involves choosing the least-privileged Copilot interaction mode for the requested work.
D:Generated code requires independently defined acceptance tests and review against the real program contract. Preventing this failure involves choosing the least-privileged Copilot interaction mode for the requested work. The observed condition is use a bounded multi-file editing workflow and manually review all proposed edits rather than authorizing agent tool execution. The prior failure was: An.
Question 171
A team reports: A developer connects an MCP server to an agent to fetch build-system metadata but the server can also modify tickets and repositories. They observe: A prompt from external data causes an agent to propose an unrequested write operation. Which ONE technical decision most directly corrects the cause?
- Use a bounded multi-file editing workflow and manually review all proposed edits rather than authorizing agent tool execution.
- Use a configured custom agent with relevant instructions and tool restrictions for the review task.
- Limit MCP tool permissions and require review before actions; treat tool outputs as untrusted external context.
- Use supported IDE Agent mode with the repository, test commands, and precise acceptance criteria while supervising tool actions.
Explanation
Correct Answer
C:The MCP integration expanded the agent's data and action trust boundary.
Incorrect Answers
A:The selected workflow granted broader tools than the task required. The MCP integration expanded the agent's data and action trust boundary.
B:Specialized objectives and tool boundaries were not provided. The MCP integration expanded the agent's data and action trust boundary.
D:The task required multi-step file/tool coordination, not a single inline continuation. The MCP integration expanded the agent's data and action trust boundary.
Question 172
A team has implemented the following change: Limit MCP tool permissions and require review before actions; treat tool outputs as untrusted external context. The original requirement was: A developer connects an MCP server to an agent to fetch build-system metadata but the server can also modify tickets and repositories. Which ONE validation test best demonstrates the intended behavior?
- Inspect the entire change set and validate the requested files changed as intended.
- Inspect granted tools and try a read-only metadata query without permitting ticket changes.
- Run it against a sample auth change and inspect the reported code references without unexpected edits.
- Inspect the proposed multi-file diff and run the project's migration and test suite.
Explanation
Correct Answer
B:Inspect granted tools and try a read-only metadata query without permitting ticket changes. Verification of this result involves inspecting granted tools and try a read-only metadata query without permitting ticket changes.
Incorrect Answers
A:Inspect the entire change set and validate the requested files changed as intended. It does not inspect granted tools and try a read-only metadata query without permitting ticket changes.
C:Run it against a sample auth change and inspect the reported code references without unexpected edits. Verification of this result involves inspecting granted tools and try a read-only metadata query without permitting ticket changes.
D:Generated code requires independently defined acceptance tests and review against the real program contract. It does not inspect granted tools and try a read-only metadata query without permitting ticket changes.
Question 173
During troubleshooting, investigators observe: A prompt from external data causes an agent to propose an unrequested write operation. The system context is: A developer connects an MCP server to an agent to fetch build-system metadata but the server can also modify tickets and repositories. Which ONE cause most directly explains the observed failure?
- A missing formatter configuration caused code style changes unrelated to the external MCP tool permission boundary.
- The MCP integration expanded the agent's data and action trust boundary.
- The task required multi-step file/tool coordination, not a single inline continuation.
- Specialized objectives and tool boundaries were not provided.
Explanation
Correct Answer
B:The remedy should therefore be: Limit MCP tool permissions and require review before actions; treat tool outputs as untrusted external context.
Incorrect Answers
A:Formatter configuration cannot explain why an MCP server was allowed to read or modify repositories and tickets. The fault is an expanded external action and data boundary.
C:The task required multi-step file/tool coordination, not a single inline continuation. The MCP integration expanded the agent's data and action trust boundary.
D:Specialized objectives and tool boundaries were not provided. The MCP integration expanded the agent's data and action trust boundary.
Question 174
The current design is: Limit MCP tool permissions and require review before actions; treat tool outputs as untrusted external context. The prior failure was: A prompt from external data causes an agent to propose an unrequested write operation. Which ONE ongoing control best protects against recurrence of that specific failure?
- Keep custom agent scopes explicit and check output against human review requirements.
- Review MCP server provenance and per-tool approvals before enabling external actions.
- Check only whether the generated source code compiles, without reviewing what the MCP server can read or write.
- Choose the least-privileged Copilot interaction mode for the requested work.
Explanation
Correct Answer
B:The MCP integration expanded the agent's data and action trust boundary. Review MCP server provenance and per-tool approvals before enabling external actions.
Incorrect Answers
A:Generated code requires independently defined acceptance tests and review against the real program contract. Preventing this failure involves reviewing MCP server provenance and per-tool approvals before enabling external actions.
C:Compilation tells nothing about the external MCP server permissions or which data leave the repository. Review server provenance and precise tool scopes instead.
D:Generated code requires independently defined acceptance tests and review against the real program contract. Preventing this failure involves reviewing MCP server provenance and per-tool approvals before enabling external actions. The observed condition is limit MCP tool permissions and require review before actions; treat tool outputs as untrusted external context. The prior failure was: A prompt from external data.
Question 175
A team reports: Several repositories use one mandatory convention: all generated Python tests must run under pytest and include type annotations. They observe: Generated tests use unittest despite the repository's established pytest policy. Which ONE technical decision most directly corrects the cause?
- Store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output.
- Add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths.
- Create a reusable `.github/prompts/*.prompt.md` template rather than copy prose repeatedly into each chat.
- Install the supported GitHub Copilot extension and sign in with the licensed GitHub account in the editor.
Explanation
Correct Answer
A:The coding standard was not provided as durable repository context.
Incorrect Answers
B:Adding supported path-specific `.github/instructions/*.instructions.md` files and scoping them to relevant paths would not store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output.
C:Creating a reusable `.github/prompts/*.prompt.md` template rather than copy prose repeatedly into each chat would not store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output.
D:Installing the supported GitHub Copilot extension and sign in with the licensed GitHub account in the editor would not store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output.
Question 176
A team has implemented the following change: Store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output. The original requirement was: Several repositories use one mandatory convention: all generated Python tests must run under pytest and include type annotations. Which ONE validation test best demonstrates the intended behavior?
- Ask Copilot for a sample unit test in each repository and confirm pytest/style compliance.
- Open a permitted source file and verify Copilot is enabled and generates a test inline suggestion.
- Invoke the prompt file with a test module and confirm the expected headings and input variables are filled.
- Ask Copilot to modify one TSX file and one Go file and check each follows its own conventions.
Explanation
Correct Answer
A:Ask Copilot for a sample unit test in each repository and confirm pytest/style compliance.
Incorrect Answers
B:Generated code requires independently defined acceptance tests and review against the real program contract. Verification of this result involves ask Copilot for a sample unit test in each repository and confirm pytest/style compliance.
C:Generated code requires independently defined acceptance tests and review against the real program contract. Verification of this result involves ask Copilot for a sample unit test in each repository and confirm pytest/style compliance. The observed condition is a team has implemented the following change: Store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output. The.
D:Generated code requires independently defined acceptance tests and review against the real program contract. Verification of this result involves ask Copilot for a sample unit test in each repository and confirm pytest/style compliance. The observed condition is a team has implemented the following change: Store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output. The. Verification of this result involves ask Copilot for a sample unit test in each repository and confirm pytest/style compliance. The named alternative instead concerns ask copilot to modify one tsx file and one go file and check each follows its own conventions.
Question 177
During troubleshooting, investigators observe: Generated tests use unittest despite the repository's established pytest policy. The system context is: Several repositories use one mandatory convention: all generated Python tests must run under pytest and include type annotations. Which ONE cause most directly explains the observed failure?
- One unscoped repository-wide instruction conflated incompatible language standards.
- Task-specific prompt guidance was not stored as a reusable artifact.
- The coding standard was not provided as durable repository context.
- The local editor integration and account authentication were not configured.
Explanation
Correct Answer
C:The remedy should therefore be: Store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output.
Incorrect Answers
A:One unscoped repository-wide instruction conflated incompatible language standards. The coding standard was not provided as durable repository context.
B:Task-specific prompt guidance was not stored as a reusable artifact. The coding standard was not provided as durable repository context.
D:The local editor integration and account authentication were not configured. The missing change involves the coding standard was not provided as durable repository context.
Question 178
The current design is: Store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output. The prior failure was: Generated tests use unittest despite the repository's established pytest policy. Which ONE ongoing control best protects against recurrence of that specific failure?
- Test instruction applicability after changing directory structure or file globs.
- Keep shared instructions short, version-controlled and aligned with CI enforcement.
- Audit IDE plugin versions and authenticated account after new workstation provisioning.
- Version and test prompt files for recurring, bounded engineering workflows.
Explanation
Correct Answer
B:The coding standard was not provided as durable repository context. Keep shared instructions short, version-controlled and aligned with CI enforcement.
Incorrect Answers
A:Generated code requires independently defined acceptance tests and review against the real program contract. It would leave keep shared instructions short, version-controlled and aligned with CI enforcement without the specified safeguard.
C:Audit IDE plugin versions and authenticated account after new workstation provisioning. It would leave keep shared instructions short, version-controlled and aligned with CI enforcement without the specified safeguard.
D:Generated code requires independently defined acceptance tests and review against the real program contract. Preventing this failure involves keep shared instructions short, version-controlled and aligned with CI enforcement.
Question 179
A team reports: A monorepo has a confirmed-loaded `.github/copilot-instructions.md` that mixes React-specific rules with incompatible Go service rules; no path-specific overrides are present. They observe: Frontend practices are applied to Go code or vice versa. Which ONE technical decision most directly corrects the cause?
- Install the supported GitHub Copilot extension and sign in with the licensed GitHub account in the editor.
- Add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths.
- Store repository-wide Copilot instructions in `.github/copilot-instructions.md` and review the affected output.
- Create a reusable `.github/prompts/*.prompt.md` template rather than copy prose repeatedly into each chat.
Explanation
Correct Answer
B:One unscoped repository-wide instruction conflated incompatible language standards.
Incorrect Answers
A:Installing the supported GitHub Copilot extension and sign in with the licensed GitHub account in the editor would not add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths.
C:Storing repository-wide Copilot instructions in `.github/copilot-instructions.md` and reviewing the affected output would not add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths.
D:Creating a reusable `.github/prompts/*.prompt.md` template rather than copy prose repeatedly into each chat would not add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths.
Question 180
A team has implemented the following change: Add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths. The original requirement was: A monorepo has a confirmed-loaded `.github/copilot-instructions.md` that mixes React-specific rules with incompatible Go service rules; no path-specific overrides are present. Which ONE validation test best demonstrates the intended behavior?
- Invoke the prompt file with a test module and confirm the expected headings and input variables are filled.
- Ask Copilot for a sample unit test in each repository and confirm pytest/style compliance.
- Open a permitted source file and verify Copilot is enabled and generates a test inline suggestion.
- Ask Copilot to modify one TSX file and one Go file and check each follows its own conventions.
Explanation
Correct Answer
D:Ask Copilot to modify one TSX file and one Go file and check each follows its own conventions.
Incorrect Answers
A:Generated code requires independently defined acceptance tests and review against the real program contract. Verification of this result involves ask Copilot to modify one TSX file and one Go file and check each follows its own conventions.
B:Generated code requires independently defined acceptance tests and review against the real program contract. Verification of this result involves ask Copilot to modify one TSX file and one Go file and check each follows its own conventions. The observed condition is a team has implemented the following change: Add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths. The.
C:Generated code requires independently defined acceptance tests and review against the real program contract. Verification of this result involves ask Copilot to modify one TSX file and one Go file and check each follows its own conventions. The observed condition is a team has implemented the following change: Add supported path-specific `.github/instructions/*.instructions.md` files and scope them to relevant paths. The. Verification of this result involves ask Copilot to modify one TSX file and one Go file and check each follows its own conventions. The named alternative instead concerns open a permitted source file and verify copilot is enabled and generates a test inline suggestion.
GitHub Copilot (GH-300) Practice Tests