Anthropic CCAR-P (Claude Certified Architect - Professional) Practice Test 19

Question 180

A support application repeatedly sends an unchanged reference handbook along with short, unrelated customer questions. After changing cache placement, which measurement would demonstrate actual prefix reuse and responsiveness?

  1. Cache-read input tokens and end-to-end latency over repeated requests
  2. Task success with reduced irrelevant instruction tokens across mixed workflows
  3. Cross-application prompt-version alignment and regression pass rate
  4. Prefix reuse ratio when request metadata changes across calls

Explanation

Correct Answer

A:

Cache-hit telemetry verifies the optimization rather than relying on estimated token savings; measure cache-read input tokens and end-to-end latency over repeated requests at the release gate; put reusable instructions and reference material in a consistent prefix and use supported prompt-cache controls while placing variable requests afterward.

Incorrect Answers

B:

On-demand loading must retain quality while avoiding unnecessary context overhead; keep Skills discoverable and load task-relevant instructions or resources on demand, maintaining scope and permission constraints; this tests task success with reduced irrelevant instruction tokens across mixed workflows; the actual release condition is cache-read input tokens and end-to-end latency over repeated requests.

C:

A shared module succeeds when consuming workflows inherit the same reviewed rule set; maintain a versioned reusable prompt module with explicit parameters and regression tests across consumers; this tests cross-application prompt-version alignment and regression pass rate; the actual release condition is cache-read input tokens and end-to-end latency over repeated requests.

D:

Controlled request pairs reveal avoidable cache invalidation; move changing metadata after the stable prefix so cache keys remain reusable across otherwise similar requests; this tests prefix reuse ratio when request metadata changes across calls; the actual release condition is cache-read input tokens and end-to-end latency over repeated requests.

Question 181

An application team runs a custom search executor although a supported managed tool exists; its bespoke responses also lose source provenance. Which integration approach should the team assess instead?

  1. Use agent-to-agent exchange for independently owned specialist systems
  2. Use MCP to expose reusable tools and resources across compatible clients
  3. Prefer managed server-side tools when appropriate to the platform
  4. Use event notifications to drive state changes rather than constant polling

Explanation

Correct Answer

C:

A supported server-executed tool can offload execution to the provider while the application retains usage, privacy, and result-validation responsibilities; managed execution should reduce infrastructure burden without weakening answer grounding; verify the repair using search coverage, tool cost, and provenance checks with no application-owned execution endpoint.

Incorrect Answers

A:

Messaging between agents does not make their outputs trustworthy or authorize cross-system actions automatically; an agent-to-agent contract can exchange bounded tasks and outputs while preserving local autonomy, identity, and authority of each specialist system; that control is assessed by successful authenticated task exchange with each agent retaining its own permission checks; the incident instead requires search coverage, tool cost, and provenance checks with no application-owned execution endpoint.

B:

MCP is an interface protocol, not a substitute for access policy or secure server implementation; a maintained MCP server can standardize discoverable tool and resource contracts while each client still enforces authentication and trust; that control is assessed by cross-client contract compatibility and unauthorized-action denial for the same MCP service; the incident instead requires search coverage, tool cost, and provenance checks with no application-owned execution endpoint.

D:

Events still require deduplication, retries, and authorization checks to avoid duplicate downstream effects; an authenticated event or webhook integration can trigger bounded processing only when upstream state changes; that control is assessed by event-to-index update delay and duplicate-event idempotency under replay; the incident instead requires search coverage, tool cost, and provenance checks with no application-owned execution endpoint.

Question 182

Every ambiguous query gets the same minimal processing and causes avoidable confident billing mistakes. Which intervention targets the defect rather than its symptoms?

  1. Evaluate model choice against the real input distribution
  2. Use stronger reasoning capacity for genuinely complex decisions
  3. Use confidence-aware routing with a bounded fallback
  4. Gate model-version updates on regression tests

Explanation

Correct Answer

C:

Combine an evaluated uncertainty trigger with escalation to a stronger model or reviewer, using explicit limits and post-route quality measurement; routing is valuable only if it captures difficult cases without needlessly escalating every simple request; verify the repair using high-risk escalation recall and total cost under the required error ceiling.

Incorrect Answers

A:

Vendor benchmark rankings do not establish suitability for a domain-specific and skewed workload; representative workload evaluation must include input quality, document lengths, languages, and failure consequences before setting model policy; that control is assessed by per-segment acceptance rate across language, length, and noisy-input strata; the incident instead requires high-risk escalation recall and total cost under the required error ceiling.

B:

High reasoning capacity is wasteful if the task only extracts a known field or matches a stable label; escalate tasks with demonstrated reasoning difficulty to a capable model after measuring gains on cases involving cross-document dependencies; that control is assessed by correct resolution of cross-exhibit conflicts at the required latency and review cost; the incident instead requires high-risk escalation recall and total cost under the required error ceiling.

D:

Newer model branding alone cannot ensure identical tool-use or output behavior in an existing workflow; treat model replacements as versioned releases with task-specific eval gates, contract tests, canaries, and rollback capability; that control is assessed by no contract regressions and acceptable quality/latency deltas on the proposed release; the incident instead requires high-risk escalation recall and total cost under the required error ceiling.

Question 183

A network failure is silently interpreted as the factual absence of an active subscription. Which intervention targets the defect rather than its symptoms?

  1. Revalidate few-shot examples after policy changes
  2. Separate external tool faults from model decisions
  3. Verify whether claims are supported by supplied sources
  4. Test for model capability mismatch on complex reasoning

Explanation

Correct Answer

B:

Propagate typed tool errors, retries, and unknown states instead of converting transport failures into negative facts; fault injection distinguishes tool availability from genuine absence of data; verify the repair using correct 'unknown' handling and no false denials during injected API failures.

Incorrect Answers

A:

Index re-embedding cannot resolve a stale rule embedded directly in the prompt template; audit examples and prompt-version precedence when behavior consistently imitates outdated demonstrations; that control is assessed by correct labels on scenarios affected by the policy revision; the incident instead requires correct 'unknown' handling and no false denials during injected API failures.

C:

A well-formatted citation is not proof that the cited text supports a particular assertion; compare material claims against cited evidence and flag unsupported statements for abstention or review; that control is assessed by supported-claim rate on source-checked summaries with seeded unsupported numbers; the incident instead requires correct 'unknown' handling and no false denials during injected API failures.

D:

Rebuilding retrieval is unlikely to help when required evidence was already consistently supplied; hold evidence constant and compare validated model capabilities or staged reasoning designs on the failed complexity class; that control is assessed by correct conflict resolution with fixed retrieved evidence across candidate reasoning configurations; the incident instead requires correct 'unknown' handling and no false denials during injected API failures.

Question 184

A transient worker crash restarts all stages, including external data pulls that already succeeded. What is the most direct remediation?

  1. Partition research by source and reconcile conflicts explicitly
  2. Checkpoint long-running work for replay and recovery
  3. Specify input and output contracts for each subtask
  4. Divide work along independently verifiable artifacts
  5. Separate planning proposals from privileged execution

Explanation

Correct Answer

B:

Durable checkpoints record validated intermediate artifacts and resume at safe boundaries with idempotent side effects; replay testing verifies the precise recovery behavior the pipeline requires; verify the repair using successful recovery after injected worker failures without duplicated writes or stages.

Incorrect Answers

A:

Simple partitioning can preserve conflicting claims unless the merge step includes a reconciliation rule; analyze source partitions separately, retain provenance, and reconcile contradictions through a dedicated synthesis stage; that control is assessed by conflict detection recall and citation coverage across independently processed source sets; the incident instead requires successful recovery after injected worker failures without duplicated writes or stages.

C:

Contracts define allowed shapes and semantics but cannot ensure a source is truthful; typed versioned contracts and compatibility tests let stages evolve while preserving their callers' assumptions; that control is assessed by contract test pass rate for independently upgraded stages and invalid payload rejection; the incident instead requires successful recovery after injected worker failures without duplicated writes or stages.

D:

Artifact decomposition controls work structure but does not set the organization-wide release SLA; separate stages with typed artifacts allow validation and targeted reruns at meaningful boundaries; that control is assessed by stage-specific defect localization and repair without redoing previously validated stages; the incident instead requires successful recovery after injected worker failures without duplicated writes or stages.

E:

Planning/execution separation does not fix whether the recommended plan uses current policy data; use a planning artifact that is reviewed and handed to a distinct authorized executor; never treat proposed tool calls as permission to mutate state; that control is assessed by unauthorized mutations during plan-only tests, with a target of zero; the incident instead requires successful recovery after injected worker failures without duplicated writes or stages.

Question 185

A free-running agent sometimes skips validation and drafts a recommendation on an unverified serial number. What design correction resolves the reported failure?

  1. Use a bounded tool-using agent for exploratory investigation
  2. Embed a small agent inside a controlled workflow
  3. Use a deterministic workflow for known ordered stages
  4. Use a single augmented-LLM call for a bounded text transformation

Explanation

Correct Answer

C:

Explicit state transitions preserve required order, testing points, and deterministic branch conditions; Claude can supply judgment inside selected stages; stage-completion evidence verifies that mandatory controls were not left to agent discretion; verify the repair using rate of recommendations backed by all required completed stage checks.

Incorrect Answers

A:

Agent autonomy adds cost and uncertainty to tasks whose steps are already predictable and linear; bounded agentic planning chooses follow-up tools from observed evidence while enforcing time, budget, and tool permissions; that control is assessed by resolution of unseen diagnostic branches within a specified tool-call and time budget; the incident instead requires rate of recommendations backed by all required completed stage checks.

B:

A fully autonomous controller is not needed when most transitions and side effects are governed by policy; place bounded exploratory reasoning inside one stage of a deterministic state machine; enforce approval and write permissions outside the agent; that control is assessed by disputed-case resolution with every refund still passing the identity and approval gates; the incident instead requires rate of recommendations backed by all required completed stage checks.

D:

A direct call is poor for tasks that require unknown sequences of external discovery or system writes; for a fixed transformation, a prompt enriched with only needed context is simpler, cheaper, and easier to evaluate than an agentic loop; that control is assessed by summary fidelity and p95 completion time without any unapproved tool invocation; the incident instead requires rate of recommendations backed by all required completed stage checks.

Question 186

An invoice-coding assistant is ready for acceptance, but the audit team is concerned about a small set of tax categories with costly misclassification consequences. Which metric would assess the critical mistakes rather than answer fluency?

  1. Unauthorized API requests rejected under role, tenant, and resource attack matrices
  2. p95 time from user submission to fully validated answer under load
  3. Fully loaded cost per correctly completed document
  4. Severity-weighted policy violation incidence on a reviewed adversarial test set
  5. Critical-class precision and recall on independently verified tax-code cases

Explanation

Correct Answer

E:

Label-based metrics catch mistakes that surface-level satisfaction may conceal; measure critical-class precision and recall on independently verified tax-code cases at the release gate; measure the business decision against a labeled reference, with metrics weighted for costly errors and explicit acceptance thresholds.

Incorrect Answers

A:

Direct penetration-style checks assess the enforceable security boundary; treat backend authorization denial rate and cross-scope access tests as primary security metrics, not merely refusal wording; this tests unauthorized API requests rejected under role, tenant, and resource attack matrices; the actual release condition is critical-class precision and recall on independently verified tax-code cases.

B:

Measuring the completed workflow captures actual end-user waiting, including downstream services; report p95 or p99 end-to-end latency over representative traffic rather than relying only on averages or raw model runtime; this tests p95 time from user submission to fully validated answer under load; the actual release condition is critical-class precision and recall on independently verified tax-code cases.

C:

This metric includes rework rather than rewarding low-quality cheap drafts; relate full inference and review costs to accepted outcomes that meet the task rubric, not to requests attempted; this tests fully loaded cost per correctly completed document; the actual release condition is critical-class precision and recall on independently verified tax-code cases.

D:

Risk-weighted scoring detects rare critical violations that a single average hides; define safety incident categories, risk-weighted violation rates, and appropriate exposure denominators for adversarial and real traffic; this tests severity-weighted policy violation incidence on a reviewed adversarial test set; the actual release condition is critical-class precision and recall on independently verified tax-code cases.

Question 187

A customer platform handles both simple FAQ requests and rare high-stakes reconciliation disputes under one expensive model route. What should the solution architect recommend?

  1. Choose model and architecture tiers by case impact
  2. Use a control group to establish incremental business impact
  3. Translate response-time expectations into a percentile SLO
  4. Measure productivity after correction and escalation

Explanation

Correct Answer

A:

Route by validated task complexity and consequence, using stronger reasoning or human review where evidence justifies the incremental spend; segmented evaluation reveals whether additional reasoning spend is targeted at cases that benefit; evaluate the design with quality and total cost segmented by request risk tier under a fixed accuracy floor.

Incorrect Answers

B:

Use matched baselines, controlled rollout, or appropriate difference-in-differences analysis to isolate contribution under plausible confounders; before-after counts without a credible comparison group may misattribute unrelated changes to the model; this alternative would measure incremental resolution lift compared with a comparable non-assisted cohort; it does not establish quality and total cost segmented by request risk tier under a fixed accuracy floor for the described use case.

C:

Define a p95 or p99 target against real end-to-end requests, including retrieval and tools, rather than reporting an average model time; a percentile SLO does not alone reveal whether the answers are factually correct; this alternative would measure end-to-end p95 latency and error rate against the agreed service objective; it does not establish quality and total cost segmented by request risk tier under a fixed accuracy floor for the described use case.

D:

Measure net time saved after review, appeals, and escalations, segmented by complexity, before claiming productivity; counting drafts as completed outcomes overlooks downstream human effort and defect consequences; this alternative would measure end-to-end completed cases per paid work-hour with rework and appeal volumes; it does not establish quality and total cost segmented by request risk tier under a fixed accuracy floor for the described use case.

Question 188

A customer assistant experiment assigns users to competing response strategies for multiple sessions. The analysis team wants to prevent exposure to both arms from distorting the measured lift. Which outcome validates the split?

  1. Balanced treatment cohorts and clean user-level exposure histories
  2. Measured pre-registered experiment compliance and interval estimates at the planned sample size
  3. Time from triggered guardrail breach to verified rollback
  4. Canary threshold compliance and time to revert a seeded regression

Explanation

Correct Answer

A:

Stable assignments preserve causal interpretation for repeated sessions; measure balanced treatment cohorts and clean user-level exposure histories at the release gate; assign a stable randomized unit aligned to the behavior being measured and maintain consistent treatment during the experiment.

Incorrect Answers

B:

Sound statistical procedure distinguishes stable benefit from random variance; define primary metrics, minimal meaningful effect, sample-size method, and stopping rules before examining results; this tests pre-registered experiment compliance and interval estimates at the planned sample size; the actual release condition is balanced treatment cohorts and clean user-level exposure histories.

C:

Operational readiness includes the ability to stop unsafe experiments promptly; establish automated and human-owned rollback triggers tied to critical error and unauthorized-action indicators; this tests time from triggered guardrail breach to verified rollback; the actual release condition is balanced treatment cohorts and clean user-level exposure histories.

D:

Controlled exposure and rollback readiness limit the damage of an unexpected release; deploy to a bounded cohort while observing quality, errors, and tail latency, with predefined stop and rollback thresholds; this tests canary threshold compliance and time to revert a seeded regression; the actual release condition is balanced treatment cohorts and clean user-level exposure histories.

Question 189

An access incident reveals that no one owned the application-level control. What design correction resolves the reported failure?

  1. Clarify shared responsibility for the service boundary
  2. Define SLOs and error budgets for agreed user outcomes
  3. Set observability and reporting expectations with stakeholders
  4. Communicate release behavior changes before rollout
  5. Route user feedback into a triaged, owned backlog

Explanation

Correct Answer

A:

Define provider, platform, application, security, and business-owner responsibilities in an explicit operating model; responsibility tests prevent gaps between contractual expectations and implemented controls; verify the repair using ownership coverage for failures across model, application, data, and tool layers.

Incorrect Answers

B:

Generic uptime claims can conceal slow or unusable answer experiences; establish measurable user-facing service objectives, reporting windows, and error-budget policies tied to accountable actions; that control is assessed by sLO compliance and prescribed actions during a simulated error-budget breach; the incident instead requires ownership coverage for failures across model, application, data, and tool layers.

C:

Platform telemetry is necessary but not equivalent to business service reporting; agree on outcome-aligned service reports, cadence, escalation thresholds, and audiences during solution definition; that control is assessed by delivery of agreed business and risk indicators on schedule with decision-ready interpretation; the incident instead requires ownership coverage for failures across model, application, data, and tool layers.

D:

Quiet deployment may reduce coordination overhead but damages trust when workflows or permissions change; define release notes, behavior-change notices, affected audiences, and support channels proportional to the deployment's impact; that control is assessed by affected-user awareness and support-ticket volume following staged changes; the incident instead requires ownership coverage for failures across model, application, data, and tool layers.

E:

Unstructured rating totals cannot by themselves direct a safe corrective action; classify feedback by severity and root-cause type, assign owners, link trace IDs, and feed validated cases into regression tests; that control is assessed by time from actionable feedback to triage, owner assignment, and verified fix; the incident instead requires ownership coverage for failures across model, application, data, and tool layers.

Question 190

An expense assistant receives PDFs, spreadsheets, and photographs with inconsistent field names, units, and missing metadata. What action best fits the business and technical boundary?

  1. Normalize incoming documents into a validated input contract
  2. Check model output against a business schema before side effects
  3. Use a queue and asynchronous workers for bursty workloads
  4. Persist workflow state and make retries idempotent

Explanation

Correct Answer

A:

An input-adapter layer validates document type, field types, provenance, and required metadata before the model reasons over values; a schema-conformance gate reveals extraction failures before they can contaminate reasoning; evaluate the design with percentage of ingested documents meeting schema checks before any Claude call.

Incorrect Answers

B:

Use structured output parsing, semantic validation, and a separate authorized command handler so generated text is never treated directly as a write instruction; output parsing constrains generated data but does not by itself manage retry idempotency or transport durability; this alternative would measure invalid write payloads rejected before execution, with zero partial supplier mutations; it does not establish percentage of ingested documents meeting schema checks before any Claude call for the described use case.

C:

Decouple intake from processing through durable queues, bounded workers, and status callbacks rather than keeping user connections open; queueing absorbs bursts but does not improve a weak answer rubric or justify a higher-privilege tool set; this alternative would measure job completion rate and queue age during the peak intake window; it does not establish percentage of ingested documents meeting schema checks before any Claude call for the described use case.

D:

Persist state transitions and use idempotency keys at side-effect boundaries to avoid duplicate transactions across replays; durable-state control prevents duplicate effects but cannot decide whether a refund is substantively authorized; this alternative would measure repeated and interrupted test runs yielding exactly one committed credit per approved case; it does not establish percentage of ingested documents meeting schema checks before any Claude call for the described use case.

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