Question 370
A customer complains about AI-generated information but support has no route to the original evidence and decision trace. Which design choice best matches the stated constraint?
- Triage incidents into versioned regression cases with prioritized remediation.
- Add operational acceptance gates before calling it production-ready.
- Use prioritized intake categories with owners, severity and response SLAs.
- Implement a governed complaint and trace-reconstruction process.
Explanation
Correct Answer
D:Operational learning and accountability require linking user reports to the responsible system state. Assign operational ownership of audit-supported customer remediation and exercise fallback or rollback during staged deployment, not only after a serious incident.
Incorrect Answers
A:Improving incident-driven evaluation improvement does not imply audit-supported customer remediation. The selected design must implement a governed complaint and trace-reconstruction process, whereas this choice omits that direct protection or verification step.
B:This recommendation targets production readiness beyond proof of concept at a different point in the workflow. It leaves audit-supported customer remediation outside the design contract and does not require the team to implement a governed complaint and trace-reconstruction process.
C:This is a valid architectural concern when optimizing structured post-launch feedback triage. Here it leaves audit-supported customer remediation unaddressed, whereas the intended change must implement a governed complaint and trace-reconstruction process before the case can be considered resolved.
Question 371
A planned model upgrade fails a safety threshold for a vulnerable user cohort. Meanwhile an upstream vendor announces a breaking connector version, and the business benefit of the running assistant has declined enough that retirement may now be justified. Which THREE lifecycle decisions should owners make? Choose THREE.
- Block or roll back release based on segment-specific nonnegotiable thresholds.
- Advance the model after a canary meets its global average score, even though the vulnerable-cohort safety gate remains unmet.
- Add a general readiness checklist while leaving the announced connector deprecation and loss of business value unowned.
- Track the connector deprecation in an owned dependency register and complete compatibility migration tests before cutover.
- Reassess realized value against the operating baseline and define a pivot-or-retire decision threshold.
Explanation
Correct Answers
A:A model upgrade that worsens safety for a vulnerable cohort must not ship merely because global quality improves. Cohort-specific release gates and rollback preserve the risk budget while owners separately manage vendor deprecation and business-value reassessment.
D:A known vendor deadline becomes a preventable outage if nobody owns the version transition. An inventory, migration owner and test gate provide time to upgrade or design a supported fallback.
E:Lifecycle governance includes knowing when a formerly useful assistant is no longer worth its cost and risk. A documented decision threshold enables controlled retirement rather than indefinite operation by inertia.
Incorrect Answers
B:A high overall score cannot waive the failing cohort’s release threshold. Launch should be stopped or reversed on that segment-specific evidence irrespective of aggregate canary performance.
C:A generic launch checklist does not manage a dated vendor migration or decide whether a declining service should continue. Those require explicit dependency and lifecycle-value decisions.
Question 372
An enterprise starts collecting feedback but cannot distinguish feature requests from safety incidents. Which approach should be recommended?
- Triage incidents into versioned regression cases with prioritized remediation.
- Implement a governed complaint and trace-reconstruction process.
- Assign operating owners, on-call escalation and maintenance budgets.
- Use prioritized intake categories with owners, severity and response SLAs.
Explanation
Correct Answer
D:Feedback management must separate optional enhancements from operational threats. Assign operational ownership of structured post-launch feedback triage and exercise fallback or rollback during staged deployment, not only after a serious incident.
Incorrect Answers
A:That choice would be responsive to incident-driven evaluation improvement, but the reported defect is different. It leaves structured post-launch feedback triage unproved because there is no commitment to use prioritized intake categories with owners, severity and response SLAs.
B:There is a legitimate use case for audit-supported customer remediation, but it does not supply the test or enforcement needed for structured post-launch feedback triage. The design needs to use prioritized intake categories with owners, severity and response SLAs.
C:Consider its actual operating scope: post-launch operational ownership. The requested assurance is structured post-launch feedback triage, which calls for an implementation that can use prioritized intake categories with owners, severity and response SLAs, not merely for a related improvement.
Question 373
A development team wants consistent review conventions across repositories but every engineer writes personal instructions. What is the best production design response?
- Scope tool discovery and credentials by project role and environment.
- Run a repository-specific regression benchmark before promoting the model in shared tooling.
- Enforce tests and status checks in the CI merge policy.
- Check in reviewed project-level Claude Code guidance and maintain it with code review.
Explanation
Correct Answer
D:Shared, versioned instructions give a reliable starting convention without confusing guidance with enforced permissions.
Incorrect Answers
A:This would make sense as an additional project activity for role- and environment-scoped tool access. The priority required here is team-wide versioned coding instructions, so the workflow still needs a control that will check in reviewed project-level Claude Code guidance and maintain it with code review.
B:By itself, this improves measured coding-tool upgrades, but the case is not asking the architect to optimize that dimension. The requirement is team-wide versioned coding instructions, supported by an approach to check in reviewed project-level Claude Code guidance and maintain it with code review.
C:Its main effect concerns enforceable build acceptance rules. The case explicitly hinges on team-wide versioned coding instructions; to resolve it reliably, the implementation has to check in reviewed project-level Claude Code guidance and maintain it with code review, which this choice does not establish.
Question 374
A coding assistant can execute shell commands in CI with the same broad credentials used by administrators. Which change most directly satisfies this requirement?
- Collect redacted command, diff, test and approval traces with correlation identifiers.
- Run the assistant in a sandbox with least-privilege service identities and command limits.
- Configure targeted repository navigation and context selection rules.
- Scope tool discovery and credentials by project role and environment.
Explanation
Correct Answer
B:Execution security comes from environment controls, not from asking the assistant to avoid destructive commands.
Incorrect Answers
A:Although auditable coding-agent activity can be valuable, it is not the acceptance condition in the question. The missing provision is restricted automation execution authority, addressed by the control to run the assistant in a sandbox with least-privilege service identities and command limits.
C:It emphasizes repository-aware context selection without closing the stated gap in restricted automation execution authority. By contrast, the remedy to run the assistant in a sandbox with least-privilege service identities and command limits acts on the condition that makes this scenario fail.
D:That is primarily a technique for role- and environment-scoped tool access. It lacks the explicit measure for restricted automation execution authority that is necessary here, namely a system that will run the assistant in a sandbox with least-privilege service identities and command limits.
Question 375
A repository requires unit tests on every merged AI-assisted change, but the team relies on reminder text. Which solution best resolves this operational gap?
- Run a repository-specific regression benchmark before promoting the model in shared tooling.
- Remove secrets from prompt files and inject short-lived credentials through a vault-controlled runner.
- Check in reviewed project-level Claude Code guidance and maintain it with code review.
- Enforce tests and status checks in the CI merge policy.
Explanation
Correct Answer
D:Guidance helps developers but only branch and pipeline controls can mandate quality gates. Verify enforceable build acceptance rules with actual repository access and CI controls; host-side controls are needed to guarantee enforceable build acceptance rules.
Incorrect Answers
A:The alternative could reduce risk associated with measured coding-tool upgrades, yet the described behavior requires enforceable build acceptance rules. That outcome depends on the capability to enforce tests and status checks in the CI merge policy.
B:The proposed measure is aimed at secure credentials for ai-assisted development, not enforceable build acceptance rules. The architecture must instead enforce tests and status checks in the CI merge policy; a separate change is still needed for enforceable build acceptance rules.
C:The relevant trade-off is not primarily team-wide versioned coding instructions. A successful fix needs enforceable build acceptance rules; otherwise enforceable build acceptance rules remains unverified even after this change.
Question 376
A team uses Claude Code across several repositories. A global command allowlist grants maintenance scripts in one low-risk repository the same shell authority in a payment-service repository. What configuration change is safest?
- Use the same global allowlist but add a warning paragraph to the shared instructions.
- Separate repository-specific command permissions and approval policies, with stricter defaults for the payment codebase.
- Grant more commands in the payment repository so both repositories have equal developer convenience.
- Disable CI checks and rely on developers to notice questionable shell commands afterward.
- Keep all commands available and rely entirely on after-the-fact audit logs.
- Run the sensitive commands through a local script with no host-side authorization checks.
Explanation
Correct Answer
B:Tool permissions should reflect each repository’s risk and workflow. Project-scoped controls and explicit approvals prevent a convenience rule for maintenance scripts from becoming broad authority in a sensitive codebase.
Incorrect Answers
A:Documentation can discourage unsafe use but does not remove host execution capability. The high-risk repository requires enforceable command restrictions rather than another nonbinding warning.
C:Equalizing access by expansion increases the consequence of an unexpected or malicious command. Least privilege requires narrowing privileges to the operations each repository actually needs.
D:Post hoc attention does not bound an agent’s command authority and bypasses independent enforcement. Review gates and repository-scoped execution controls must remain in place before sensitive actions.
E:Logs can support incident investigation but do not prevent a dangerous command from executing. The payment-service repository needs preventive enforcement of narrower command permissions.
F:Relocating the command path does not constrain capabilities. A helper process inherits real host permissions unless the runner or gateway enforces a restricted approved command set.
Question 377
A code agent accesses internal deployment tools that most contributors should not see. Which intervention most directly addresses the failure?
- Remove secrets from prompt files and inject short-lived credentials through a vault-controlled runner.
- Scope tool discovery and credentials by project role and environment.
- Run the assistant in a sandbox with least-privilege service identities and command limits.
- Collect redacted command, diff, test and approval traces with correlation identifiers.
Explanation
Correct Answer
B:Developer convenience cannot justify exposing privileged production actions to every worker. Verify role- and environment-scoped tool access with actual repository access and CI controls; host-side controls are needed to guarantee role- and environment-scoped tool access.
Incorrect Answers
A:That is primarily a technique for secure credentials for ai-assisted development. It lacks the explicit measure for role- and environment-scoped tool access that is necessary here, namely a system that will scope tool discovery and credentials by project role and environment.
C:This recommendation targets restricted automation execution authority at a different point in the workflow. It leaves role- and environment-scoped tool access outside the design contract and does not require the team to scope tool discovery and credentials by project role and environment.
D:This is a valid architectural concern when optimizing auditable coding-agent activity. Here it leaves role- and environment-scoped tool access unaddressed, whereas the intended change must scope tool discovery and credentials by project role and environment before the case can be considered resolved.
Question 378
A large monorepo floods the assistant with irrelevant source files before each coding task. What should the solution architect prioritize?
- Check in reviewed project-level Claude Code guidance and maintain it with code review.
- Run a repository-specific regression benchmark before promoting the model in shared tooling.
- Configure targeted repository navigation and context selection rules.
- Enforce tests and status checks in the CI merge policy.
- Require diff review and restrict edit scope before sensitive file changes.
Explanation
Correct Answer
C:Focused code context improves signal per token and reduces irrelevant edits. Verify repository-aware context selection with actual repository access and CI controls; host-side controls are needed to guarantee repository-aware context selection.
Incorrect Answers
A:The control would be justified if team-wide versioned coding instructions were the limiting factor. In this scenario the decisive target is repository-aware context selection; success requires an architecture that will configure targeted repository navigation and context selection rules.
B:The option belongs to the broader work on measured coding-tool upgrades, but it cannot by itself deliver repository-aware context selection. A decision on this case needs the team to configure targeted repository navigation and context selection rules and verify that outcome.
D:For a workflow whose primary weakness was enforceable build acceptance rules, that proposal could be reasonable. Here the unmet obligation is repository-aware context selection, and that requires the system to configure targeted repository navigation and context selection rules.
E:Although controlled edit permissions for sensitive files can be valuable, it is not the acceptance condition in the question. The missing provision is repository-aware context selection, addressed by the control to configure targeted repository navigation and context selection rules.
Question 379
A coding assistant's tool calls are hard to reconstruct after a production incident. Which architecture adjustment is most appropriate?
- Collect redacted command, diff, test and approval traces with correlation identifiers.
- Run a repository-specific regression benchmark before promoting the model in shared tooling.
- Require diff review and restrict edit scope before sensitive file changes.
- Scope tool discovery and credentials by project role and environment.
Explanation
Correct Answer
A:Observable developer automation allows incidents to be investigated without excessive raw secret capture. Verify auditable coding-agent activity with actual repository access and CI controls; host-side controls are needed to guarantee auditable coding-agent activity.
Incorrect Answers
B:That choice would be responsive to measured coding-tool upgrades, but the reported defect is different. It leaves auditable coding-agent activity unproved because there is no commitment to collect redacted command, diff, test and approval traces with correlation identifiers.
C:There is a legitimate use case for controlled edit permissions for sensitive files, but it does not supply the test or enforcement needed for auditable coding-agent activity. The design needs to collect redacted command, diff, test and approval traces with correlation identifiers.
D:Consider its actual operating scope: role- and environment-scoped tool access. The requested assurance is auditable coding-agent activity, which calls for an implementation that can collect redacted command, diff, test and approval traces with correlation identifiers, not merely for a related improvement.
Question 380
A developer's assistant silently updates lockfiles and deploy scripts while solving a documentation typo. Which next step is most defensible?
- Remove secrets from prompt files and inject short-lived credentials through a vault-controlled runner.
- Collect redacted command, diff, test and approval traces with correlation identifiers.
- Configure targeted repository navigation and context selection rules.
- Require diff review and restrict edit scope before sensitive file changes.
Explanation
Correct Answer
D:Human-visible change boundaries catch collateral modifications that a model may introduce. Verify controlled edit permissions for sensitive files with actual repository access and CI controls; host-side controls are needed to guarantee controlled edit permissions for sensitive files.
Incorrect Answers
A:This is a valid architectural concern when optimizing secure credentials for ai-assisted development. Here it leaves controlled edit permissions for sensitive files unaddressed, whereas the intended change must require diff review and restrict edit scope before sensitive file changes before the case can be considered resolved.
B:This would make sense as an additional project activity for auditable coding-agent activity. The priority required here is controlled edit permissions for sensitive files, so the workflow still needs a control that will require diff review and restrict edit scope before sensitive file changes.
C:By itself, this improves repository-aware context selection, but the case is not asking the architect to optimize that dimension. The requirement is controlled edit permissions for sensitive files, supported by an approach to require diff review and restrict edit scope before sensitive file changes.