Exam Bundle

A comparison prompt should request decision evidence, not only a recommendation. That recommendation is relevant to a different task: A teammate needs a structured design choice between two database options. In this case the main need is: A new developer wants Copilot to write a routine from an explicit textual requirement only.

Few-shot examples help constrain desired behavior and formatting. That recommendation is relevant to a different task: An engineer wants a function generated from a specification but has no sample calls. In this case the main need is: A new developer wants Copilot to write a routine from an explicit textual requirement only.

Explicit negative constraints prevent unintended interface changes. That recommendation is relevant to a different task: A function must remain backward compatible but the initial prompt omits compatibility rules. In this case the main need is: A new developer wants Copilot to write a routine from an explicit textual requirement only.

“Latency improves while audit events remain complete and queryable.” does not demonstrate the result required here: Benchmark scripts reproduce the claimed workload comparison.

“The candidate can explain the corrected behavior and reproduce it unaided.” does not demonstrate the result required here: Benchmark scripts reproduce the claimed workload comparison.

“Representative multilingual cases meet target error rates.” does not demonstrate the result required here: Benchmark scripts reproduce the claimed workload comparison.

“Discount tests cover the zero, threshold, and upper-bound cases.” does not demonstrate the result required here: Stress tests produce correct totals without data-race findings.

“The token is revoked and the repository scan reports no usable secret.” does not demonstrate the result required here: Stress tests produce correct totals without data-race findings.

“Approved origins succeed and disallowed origins cannot read responses.” does not demonstrate the result required here: Stress tests produce correct totals without data-race findings.

Mixed-private organizational context may prevent some cloud usage and cause access surprises. That recommendation is relevant to a different task: A shared context Space includes references from different organizations with incompatible access. In this case the main need is: A support team needs Copilot to answer based on a current issue, PR, and ADR together.

Spaces ground questions in curated persistent project context. That recommendation is relevant to a different task: An engineering group shares architecture notes, issues, and repository context for repeated questions. In this case the main need is: A support team needs Copilot to answer based on a current issue, PR, and ADR together.

Summaries help reviewers understand change scope but may omit details. That recommendation is relevant to a different task: A PR author needs a concise explanation of a complex diff for reviewers. In this case the main need is: A support team needs Copilot to answer based on a current issue, PR, and ADR together.

Documented help reflects installed version capabilities. That recommendation is relevant to a different task: An engineer needs to find the available CLI slash commands without memorizing them. In this case the main need is: A developer invokes Copilot CLI in the home directory, accidentally expanding project scope.

Explicit file context focuses analysis on the relevant artifact. That recommendation is relevant to a different task: A developer wants to include a specific config file as CLI context. In this case the main need is: A developer invokes Copilot CLI in the home directory, accidentally expanding project scope.

Repository trust affects which local context and tools the assistant may access. That recommendation is relevant to a different task: A terminal session asks whether the current repository may be trusted by the assistant. In this case the main need is: A developer invokes Copilot CLI in the home directory, accidentally expanding project scope.

“The reviewed diff includes intended modules and preserves unrelated files.” does not demonstrate the result required here: The proposed expression appears as an inline suggestion and is accepted deliberately.

“The response correctly references the function's inputs and side effects.” does not demonstrate the result required here: The proposed expression appears as an inline suggestion and is accepted deliberately.

“The patch changes the intended method and tests cover both implementations.” does not demonstrate the result required here: The proposed expression appears as an inline suggestion and is accepted deliberately.

“The final commit message describes the changed behavior and tests.” does not demonstrate the result required here: Known input records produce equal results in both implementations.

“All endpoints retain equivalent error behavior after extraction.” does not demonstrate the result required here: Known input records produce equal results in both implementations.

“The proposed call-site list matches a source search.” does not demonstrate the result required here: Known input records produce equal results in both implementations.

Infrastructure text can grant more access than intended. That recommendation is relevant to a different task: An engineer adds Copilot-generated infrastructure YAML with unexpected wildcard permissions. In this case the main need is: A team plans to release an AI-generated authentication library fork without code ownership review.

Some regex constructs can trigger excessive backtracking. That recommendation is relevant to a different task: A developer accepts an AI-generated regular expression without worst-case evaluation. In this case the main need is: A team plans to release an AI-generated authentication library fork without code ownership review.

Functional validation needs expected failure behavior as well as success. That recommendation is relevant to a different task: A generated input validator passes a happy-path test but its null behavior is unspecified. In this case the main need is: A team plans to release an AI-generated authentication library fork without code ownership review.

“The patch changes the intended method and tests cover both implementations.” does not demonstrate the result required here: The extension recognizes the intended language and produces syntactically relevant code.

“The executed command matches the approved one and leaves unrelated files untouched.” does not demonstrate the result required here: The extension recognizes the intended language and produces syntactically relevant code.

“No project file is changed until the team approves a subsequent edit.” does not demonstrate the result required here: The extension recognizes the intended language and produces syntactically relevant code.

“The corrected resource resolves to the documented product page.” does not demonstrate the result required here: The corrected review no longer incorporates restricted content.

“The allowed request succeeds with safety protections intact.” does not demonstrate the result required here: The corrected review no longer incorporates restricted content.

“Sensitive generated code receives an independent review regardless of badge presence.” does not demonstrate the result required here: The corrected review no longer incorporates restricted content.

“The review record contains the matched source URL and reported license.” does not demonstrate the result required here: The committed change contains no embedded credential and scans pass.

“Explicit attack fixtures and human review supplement AI-generated checks.” does not demonstrate the result required here: The committed change contains no embedded credential and scans pass.

“The final adopted code has reviewed licensing or a clean independent replacement.” does not demonstrate the result required here: The committed change contains no embedded credential and scans pass.