Anthropic CCAR-P (Claude Certified Architect - Professional) Practice Test 60

Question 590

After installing Separation of duties for sensitive changes in a manufacturing quality workflow, the team must validate whether it has reduced the exposure that self-approval eliminates independent control over privileged operations. Which measurement is most relevant?

  1. Audit self-approved mutation attempts prevented
  2. Audit unauthorized external dispatch rate during staging tests
  3. Audit unapproved action count during simulated reviewer outages
  4. Audit reviewer detection rate of deliberately injected errors
  5. Audit detected error rate and audit sampling coverage

Explanation

Correct Answer

A:

An appropriate gate measures self-approved mutation attempts prevented. It tests the mechanism in which the control layer can require a distinct authorized human or system role for approval, rather than a cosmetic demonstration. That distinction matters because self-approval eliminates independent control over privileged operations.

Incorrect Answers

B:

When a workflow can either draft a notice or send it to all customers, the control layer can require stronger approval before externally visible or irreversible steps; however, that proxy would not demonstrate self-approved mutation attempts prevented for the selected intervention.

C:

The proposed measure would suit a setting where approval queues may be unattended overnight, since the control layer can hold or safely reject high-impact actions until review is complete; but the acceptance gate must instead inspect self-approved mutation attempts prevented to detect the targeted failure.

D:

The proposed measure would suit a setting where approvers receive model recommendations without underlying sources, since the control layer can present retrieved evidence, uncertainty and proposed effect alongside each action; by contrast, the manufacturing quality release needs evidence from self-approved mutation attempts prevented.

E:

The alternative indicator checks whether the control layer can sample approved outputs for independent periodic quality review when fully automated classifications still contain subtle long-tail errors; however, that proxy would not demonstrate self-approved mutation attempts prevented for the selected intervention.

Question 591

After a model prompt change, the casework team skips the test suite for bugs fixed in earlier versions. Which release risk does a regression gate address that user feedback alone cannot?

  1. An exposure in which a positive global mean can conceal subgroup harm
  2. An exposure in which full rollout hides regressions until broad exposure
  3. An exposure in which before-after comparisons can confound seasonality and traffic mix
  4. An exposure in which unrelated traffic differences can masquerade as improvement
  5. An exposure in which user anecdotes cannot guarantee a former defect stays resolved

Explanation

Correct Answer

E:

The loss of the control permits the situation where user anecdotes cannot guarantee a former defect stays resolved. Its governing mechanism is that the evaluation process can run a pinned critical-case suite before permitting a release; without an equivalent mechanism, a favorable demonstration cannot establish safety or quality.

Incorrect Answers

A:

If a variant helps English requests but harms low-resource-language cases, a positive global mean can conceal subgroup harm; however, the named safeguard addresses the distinct hazard that user anecdotes cannot guarantee a former defect stays resolved.

B:

When a new model configuration has limited high-risk production evidence, full rollout hides regressions until broad exposure; whereas removing Release regression gate primarily exposes the risk that user anecdotes cannot guarantee a former defect stays resolved.

C:

When a new instruction template claims to reduce escalations, before-after comparisons can confound seasonality and traffic mix; but this is not the immediate loss associated with omitting Release regression gate: user anecdotes cannot guarantee a former defect stays resolved.

D:

The competing hazard is that unrelated traffic differences can masquerade as improvement in a workload where candidate retrieval changes should be compared against identical inputs; however, the named safeguard addresses the distinct hazard that user anecdotes cannot guarantee a former defect stays resolved.

Question 592

A design authority for financial operations has established this requirement: users need automatic replies except when evidence is weak or risk is high. Which intervention should be prioritized?

  1. Implement Retrieval-augmented response pattern
  2. Implement Batch offline processing
  3. Implement Single-turn structured extraction
  4. Implement Rule engine before LLM routing
  5. Implement Human escalation at uncertainty boundary

Explanation

Correct Answer

E:

Under the stated constraints, confidence and policy triggers send ambiguous cases to a qualified reviewer. That directly addresses the operational failure without relying on a generic prompt reminder. Production depends on review capacity and calibrated abstention conditions; the result to monitor is unsafe auto-resolution rate at each escalation threshold.

Incorrect Answers

A:

When answers must reflect changing internal policies without retraining a model, the architecture can retrieve authorized current source passages before generation and attach provenance; the stated requirement remains unmet: confidence and policy triggers send ambiguous cases to a qualified reviewer.

B:

If overnight document enrichment has no conversational latency requirement, the architecture can queue and batch eligible tasks rather than forcing interactive synchronous calls; whereas the financial operations case must guard against the outcome that a forced final answer may present unsupported certainty to users.

C:

When each claim form can be parsed independently into predetermined fields, one bounded model call extracts fields without orchestration overhead; this rollout still needs the following guarantee: confidence and policy triggers send ambiguous cases to a qualified reviewer.

D:

If clear transaction thresholds determine whether a case needs model interpretation, deterministic rules cheaply dispatch known cases and reserve Claude for ambiguous ones; the stated requirement remains unmet: confidence and policy triggers send ambiguous cases to a qualified reviewer.

Question 593

During an acceptance review in energy operations, End-to-end dependency map is operational but its benefit has not been measured. Auditors are concerned that a model-only dashboard cannot explain delays in network and storage hops. Which evidence most directly settles the question?

  1. Evaluate leakage findings in log sampling and telemetry utility
  2. Evaluate bottleneck attribution coverage across major call paths
  3. Evaluate complete trace rate and untraced tool invocations
  4. Evaluate version-specific quality and latency trend clarity

Explanation

Correct Answer

B:

The correct signal is bottleneck attribution coverage across major call paths; it must be computed for representative workloads and compared with a baseline. It checks whether the integration can map the sequence of retrieval, orchestration and response assembly while surfacing the exposure that a model-only dashboard cannot explain delays in network and storage hops.

Incorrect Answers

A:

The alternative indicator checks whether the integration can emit structured metrics and redacted traces with restricted access when production traces can contain sensitive customer prompts and tool payloads; by contrast, the energy operations release needs evidence from bottleneck attribution coverage across major call paths.

C:

Where an agent call spans model requests, MCP servers and downstream APIs, a relevant check would establish whether the integration can propagate stable trace identifiers across each hop and tool invocation; but the acceptance gate must instead inspect bottleneck attribution coverage across major call paths to detect the targeted failure.

D:

When performance degrades after a template rollout, the integration can attach version identifiers to every inference and evaluation event; by contrast, the energy operations release needs evidence from bottleneck attribution coverage across major call paths.

Question 594

A program architect at media licensing has endorsed Reviewed automated code documentation. The proposed technical effect is that the engineering workflow can verify generated docs against source and running contracts. Which implementation condition is necessary to realize that effect?

  1. Establish reliable test suite and human code review
  2. Establish fixed review rubrics and deduplicated findings
  3. Establish issue reproduction, scoped permissions and merge protection
  4. Establish source-control workflow and required review metadata
  5. Establish authoritative interfaces and review ownership

Explanation

Correct Answer

E:

Readiness depends on authoritative interfaces and review ownership, not just a design approval. That condition supports the mechanism in which the engineering workflow can verify generated docs against source and running contracts. Without it, language fluency does not ensure generated documentation matches behavior; monitoring documentation contract mismatch count helps verify the deployed result.

Incorrect Answers

A:

Where agents generate plausible code that breaks existing behavior, the engineering workflow can require narrow changes with executable tests and review before merge; for this release, the team must first verify authoritative interfaces and review ownership.

B:

The setup is useful to enable the alternative's behavior where the engineering workflow can run security, maintainability and API compatibility reviews separately and multiple quality concerns need independent assessment of a large change; however, the named intervention needs authoritative interfaces and review ownership before its protection can be enforced.

C:

This prerequisite supports a method used when routine low-risk bugs have reproducible fixtures, because the engineering workflow can allow an agent to propose a patch and tests within a restricted repository scope; but the production dependency for Reviewed automated code documentation is authoritative interfaces and review ownership.

D:

This prerequisite supports a method used when release teams cannot tell why generated modifications were proposed, because the engineering workflow can attach issue links, tests, diff summaries and reviewer approvals to patches; for this release, the team must first verify authoritative interfaces and review ownership.

Question 595

A telecommunications support incident exposes a specific weakness: a free-form summary can omit a required category even when fluent. Which architectural correction addresses the cause rather than masking the symptom?

  1. Prioritize Zero-shot baseline
  2. Prioritize Structured intermediate checklist
  3. Prioritize Decomposed multi-pass reasoning
  4. Prioritize Self-consistency sampling with adjudication

Explanation

Correct Answer

B:

This intervention addresses the failure at its source because the model configuration can require a concise explicit checklist of required criteria before final output. To keep the issue from recurring, establish an auditable rubric and output slots and verify required-criterion coverage per reviewed response; otherwise a free-form summary can omit a required category even when fluent can persist.

Incorrect Answers

A:

When examples can add maintenance cost without measurable gain, the model configuration can start with concise instructions before adding demonstrations that consume context; yet it would not correct the observed weakness: a free-form summary can omit a required category even when fluent.

C:

The alternative reduces the risk that one end-to-end generation obscures which subreasoning step failed because the model configuration can compute verifiable components separately before final synthesis; however, the telecommunications support defect follows the path in which a free-form summary can omit a required category even when fluent.

D:

Although the alternative reduces the risk that repeating generations without a judging rubric can simply increase cost because the model configuration can evaluate multiple bounded candidates and adjudicate against evidence; yet it would not correct the observed weakness: a free-form summary can omit a required category even when fluent.

Question 596

The healthcare operations staging review finds that anecdotal messages without context are hard to triage. Which design intervention is the most appropriate corrective action?

  1. Introduce Structured feedback intake
  2. Introduce Escalation matrix for model incidents
  3. Introduce Service-tier differentiation
  4. Introduce Expectation calibration at pilot kickoff

Explanation

Correct Answer

A:

The engineering fix changes the mechanism rather than the presentation: the delivery team can capture task type, evidence, severity and version in one feedback queue. It requires standard feedback schema and privacy controls; a regression test should examine actionable bug reports and time to reproduce in cases where anecdotal messages without context are hard to triage.

Incorrect Answers

B:

Where users do not know whom to call when generated advice causes harm, the delivery team can map severity levels to operational, security and business response roles; whereas the healthcare operations defect follows the path in which anecdotal messages without context are hard to triage.

C:

Where internal analytics and customer-facing support have incompatible latency demands, the delivery team can commit separate targets and fallback behaviors by priority class; yet it would not correct the observed weakness: anecdotal messages without context are hard to triage.

D:

Although the alternative reduces the risk that overselling demo accuracy creates unsupported production expectations because the delivery team can communicate pilot limitations, confidence bounds and human review scope; but this remedy operates on another boundary, not the case where anecdotal messages without context are hard to triage.

Question 597

An independent assessment of a healthcare operations Claude system reports that self-approval eliminates independent control over privileged operations. Which change best closes the exposed control gap?

  1. Prioritize Separation of duties for sensitive changes
  2. Prioritize Random audit of low-risk automation
  3. Prioritize Escalation timeout and fail-closed route
  4. Prioritize Risk-tiered approval workflow
  5. Prioritize Override reason capture

Explanation

Correct Answer

A:

The reported failure calls for the control under which the control layer can require a distinct authorized human or system role for approval. A prompt-only reminder or unrelated metric would not address the risk that self-approval eliminates independent control over privileged operations. Evidence of resolution requires identity-separated approval roles and tamper-evident logs and an improvement in self-approved mutation attempts prevented.

Incorrect Answers

B:

Although the alternative reduces the risk that no-review automation can silently degrade after a distribution shift because the control layer can sample approved outputs for independent periodic quality review; yet it would not correct the observed weakness: self-approval eliminates independent control over privileged operations.

C:

This addresses the failure where automatic execution after a timeout defeats the approval requirement, because the control layer can hold or safely reject high-impact actions until review is complete; but this remedy operates on another boundary, not the case where self-approval eliminates independent control over privileged operations.

D:

When blanket approval for every step wastes capacity; blanket autonomy hides risk, the control layer can send consequential actions to authorized reviewers while allowing bounded automation elsewhere; whereas the healthcare operations defect follows the path in which self-approval eliminates independent control over privileged operations.

E:

When raw accept-or-reject clicks cannot isolate model or retrieval defects, the control layer can capture structured reasons and attach them to evaluation feedback; yet it would not correct the observed weakness: self-approval eliminates independent control over privileged operations.

Question 598

A design authority for internal audit has established this requirement: generated answers omit a rule even though it exists in the knowledge base. Which intervention should be prioritized?

  1. Unsupported-claim provenance analysis
  2. Model capability mismatch test
  3. Context truncation investigation
  4. Retrieval recall diagnosis
  5. Overfit-to-examples analysis

Explanation

Correct Answer

D:

The mechanism is specific: the evaluation process can inspect whether relevant source passages reached the model. A credible implementation requires query-to-source relevance labels, while its acceptance test monitors retrieval recall at k on missing-rule cases. The alternative approaches leave the risk that rewriting prompts cannot compensate for evidence never retrieved.

Incorrect Answers

A:

When a response confidently cites text that does not support the claim, the evaluation process can validate claim-to-span entailment rather than just link presence; it does not ensure the required behavior: the evaluation process can inspect whether relevant source passages reached the model.

B:

For a workload where complex cross-document inferences fail despite accurate retrieved passages, the evaluation process can evaluate a more capable tier on the same evidence and instructions; whereas the telecommunications support case must guard against the outcome that rewriting prompts cannot compensate for evidence never retrieved.

C:

For a workload where failures spike only when lengthy contracts include appendices, the evaluation process can inspect token budgets and the presence of decisive passages after packing; the alternative leaves the central requirement unproven: the evaluation process can inspect whether relevant source passages reached the model.

E:

If a few-shot classifier works on familiar categories but fails novel wording, the evaluation process can test whether examples bias the model to surface patterns rather than policy rules; whereas the telecommunications support case must guard against the outcome that rewriting prompts cannot compensate for evidence never retrieved.

Question 599

For a supply-chain planning rollout, the proposed design uses Human escalation at uncertainty boundary, relying on the claim that confidence and policy triggers send ambiguous cases to a qualified reviewer. What enabling control must be in place to make that design enforceable?

  1. Prepare versioned step contracts and deterministic exception routes
  2. Prepare a restricted tool catalog and explicit stop conditions
  3. Prepare indexed authoritative documents and retrieval access checks
  4. Prepare review capacity and calibrated abstention conditions

Explanation

Correct Answer

D:

What makes the control operable is review capacity and calibrated abstention conditions. It supplies the precondition for the mechanism whereby confidence and policy triggers send ambiguous cases to a qualified reviewer. The residual failure to guard against is that a forced final answer may present unsupported certainty to users, measured through unsafe auto-resolution rate at each escalation threshold.

Incorrect Answers

A:

For the separate challenge where invoice approvals must follow the same mandated checks in the same order, orchestration fixes the sequence and uses Claude only inside bounded steps; but the production dependency for Human escalation at uncertainty boundary is review capacity and calibrated abstention conditions.

B:

For the separate challenge where a service request requires choosing among several tools based on emerging evidence, a controlled agent loop selects the next tool under budgets and permission gates; for this release, the team must first verify review capacity and calibrated abstention conditions.

C:

The setup is useful to enable the alternative's behavior where the architecture can retrieve authorized current source passages before generation and attach provenance and answers must reflect changing internal policies without retraining a model; however, the named intervention needs review capacity and calibrated abstention conditions before its protection can be enforced.

Question 600

The post-incident analysis for retail operations concludes that a model-only dashboard cannot explain delays in network and storage hops. Which remediation targets the relevant failure boundary?

  1. Introduce Model-and-prompt version tracing
  2. Introduce Distributed trace correlation
  3. Introduce End-to-end dependency map
  4. Introduce Privacy-aware telemetry design

Explanation

Correct Answer

C:

The remediation acts at the relevant design boundary: the integration can map the sequence of retrieval, orchestration and response assembly. That counteracts the observed hazard that a model-only dashboard cannot explain delays in network and storage hops. Rollout should include instrumented service spans and common time reference, with regression evidence from bottleneck attribution coverage across major call paths.

Incorrect Answers

A:

When mixing versions destroys the ability to attribute a regression, the integration can attach version identifiers to every inference and evaluation event; whereas the retail operations defect follows the path in which a model-only dashboard cannot explain delays in network and storage hops.

B:

When isolated logs cannot reveal the critical latency or failure path, the integration can propagate stable trace identifiers across each hop and tool invocation; yet it would not correct the observed weakness: a model-only dashboard cannot explain delays in network and storage hops.

D:

This addresses the failure where full raw payload logging can breach retention and minimization obligations, because the integration can emit structured metrics and redacted traces with restricted access; whereas the retail operations defect follows the path in which a model-only dashboard cannot explain delays in network and storage hops.

Claude Certified Architect - Professional (CCAR-P) Practice Tests