Exam Bundle

“The sensitive file remains excluded while permitted files still receive suggestions.” does not demonstrate the result required here: A permitted account receives suggestions only after the correct access is restored.

“Suggestions work in an allowed fixture while remaining absent in the excluded path.” does not demonstrate the result required here: A permitted account receives suggestions only after the correct access is restored.

“Sensitive generated code receives an independent review regardless of badge presence.” does not demonstrate the result required here: A permitted account receives suggestions only after the correct access is restored.

Permission scoping and context isolation limit accidental cross-project use. That recommendation is relevant to a different task: A project wants to prevent code from one private repository being reused as context for another project. In this case the main need is: A private repository contains generated credential fixtures that should not inform suggestions in other files.

Excluded content should not be analyzed by supported code review surfaces. That recommendation is relevant to a different task: A repository administrator needs to exclude a sensitive fixture from Copilot code review. In this case the main need is: A private repository contains generated credential fixtures that should not inform suggestions in other files.

GitHub exclusions are not guaranteed to bind external MCP servers. That recommendation is relevant to a different task: A platform team assumes file exclusions are guaranteed for third-party MCP integrations. In this case the main need is: A private repository contains generated credential fixtures that should not inform suggestions in other files.

“Comparable tests report correctness, latency and resource use under the same inputs.” does not demonstrate the result required here: Both candidate implementations are evaluated against the same acceptance tests.

“The acceptance decision uses tests and evidence rather than style.” does not demonstrate the result required here: Both candidate implementations are evaluated against the same acceptance tests.

“A concise context restatement improves fidelity compared with the overloaded thread.” does not demonstrate the result required here: Both candidate implementations are evaluated against the same acceptance tests.

Cancellation is part of correct lifecycle management. That recommendation is relevant to a different task: A generated asynchronous call path ignores cancellation and leaves resources open. In this case the main need is: A Copilot fix passes unit tests but fails the production language version.

Authentication patches can be subtly insecure despite compiling. That recommendation is relevant to a different task: Copilot produces a patch for a broken OAuth flow with no test evidence. In this case the main need is: A Copilot fix passes unit tests but fails the production language version.

Plausible parsers can miss semantic invalidity. That recommendation is relevant to a different task: An AI-generated parser accepts an invalid date such as February 30. In this case the main need is: A Copilot fix passes unit tests but fails the production language version.

“Invalid tokens fail and a correctly scoped token succeeds.” does not demonstrate the result required here: A token missing the required claim is denied.

“Every control is reachable and usable without a mouse.” does not demonstrate the result required here: A token missing the required claim is denied.

“A revoked access rule takes effect within the required period.” does not demonstrate the result required here: A token missing the required claim is denied.

Authentication tests need positive and negative principal fixtures. That recommendation is relevant to a different task: A pull request changes a JWT authorization middleware. In this case the main need is: A payment calculation uses floating-point arithmetic and fails on rounding cases.

Adversarial inputs can reveal parser edge cases. That recommendation is relevant to a different task: A file parser should withstand malformed and oversized input without crashing. In this case the main need is: A payment calculation uses floating-point arithmetic and fails on rounding cases.

Test generation can expose negative authorization paths. That recommendation is relevant to a different task: An endpoint must reject invalid authorization tokens while still allowing valid ones. In this case the main need is: A payment calculation uses floating-point arithmetic and fails on rounding cases.

Agent session artifacts provide evidence of actual work. That recommendation is relevant to a different task: An agent has produced multiple commits in a cloud session and the user must inspect changes. In this case the main need is: A shared context Space includes references from different organizations with incompatible access.

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 shared context Space includes references from different organizations with incompatible access.

Relevant structured context limits unsupported inference. That recommendation is relevant to a different task: A support team needs Copilot to answer based on a current issue, PR, and ADR together. In this case the main need is: A shared context Space includes references from different organizations with incompatible access.

AI translation may omit critical qualifiers or distort risks. That recommendation is relevant to a different task: A generated translation of an error message changes a security warning into permissive text. In this case the main need is: A chatbot suggests downloading a debugging binary from a domain that impersonates the vendor.

Models can invent credible dependency names and versions. That recommendation is relevant to a different task: An assistant generates a package name that does not exist in the approved dependency registry. In this case the main need is: A chatbot suggests downloading a debugging binary from a domain that impersonates the vendor.

AI-generated logic can reproduce bias or overlook protected-group effects. That recommendation is relevant to a different task: A Copilot recommendation seems to favor one demographic group in a user-ranking algorithm. In this case the main need is: A chatbot suggests downloading a debugging binary from a domain that impersonates the vendor.

“Compliance review confirms required notices and attribution are present.” does not demonstrate the result required here: The accepted code has a recorded provenance review or is independently rewritten.

“Review evidence confirms approved primitives and tests cover misuse and invalid inputs.” does not demonstrate the result required here: The accepted code has a recorded provenance review or is independently rewritten.

“The query excludes unrelated personal attributes and passes documented review cases.” does not demonstrate the result required here: The accepted code has a recorded provenance review or is independently rewritten.

AI summarization can remove legally significant constraints. That recommendation is relevant to a different task: A summary of a license notice omits attribution obligations for derived code. In this case the main need is: An assistant generates a package name that does not exist in the approved dependency registry.

Models do not inherently verify adversarial resilience of their own suggestions. That recommendation is relevant to a different task: A team wants to use Copilot-generated security controls without adversarial testing. In this case the main need is: An assistant generates a package name that does not exist in the approved dependency registry.

Generated output may resemble copyrighted or licensed source material. That recommendation is relevant to a different task: An assistant returns a code snippet copied verbatim from an unfamiliar public project. In this case the main need is: An assistant generates a package name that does not exist in the approved dependency registry.

“The example starts and demonstrates the requested feature.” does not demonstrate the result required here: Generated SQL and query outputs match the baseline.

“Before-and-after fixtures produce identical outputs.” does not demonstrate the result required here: Generated SQL and query outputs match the baseline.

“Comments accurately describe each allowed transition.” does not demonstrate the result required here: Generated SQL and query outputs match the baseline.