Anthropic CCAR-P (Claude Certified Architect - Professional) Practice Test 26

Question 250

A procurement assistant summarizes external vendor portals that contain arbitrary embedded prose. Which choice most directly satisfies the documented goal?

  1. Enforce least privilege where tool actions execute
  2. Screen unsafe outputs in proportion to the use case
  3. Treat retrieved instructions as untrusted data
  4. Sandbox untrusted code and file operations
  5. Apply controls at input, tool execution, and output boundaries

Explanation

Correct Answer

C:

Keep source contents in a lower-trust evidence role and prevent embedded commands from crossing into tool authority; simulated hostile documents test the trust boundary directly; evaluate the design with attack success rate on adversarial web pages and zero unauthorized egress calls.

Incorrect Answers

A:

Enforce role and scope in the backend execution service using credentials that cannot perform the forbidden action; natural-language reminders do not remove capabilities exposed by the API or bearer token; this alternative would measure zero payment writes when tested with the analytics identity and forged instructions; it does not establish attack success rate on adversarial web pages and zero unauthorized egress calls for the described use case.

B:

Evaluate output safety against a documented risk rubric and use appropriate escalation or refusal for high-risk claims; an input-only filter cannot detect every unsafe completion, especially when harm arises from the generated detail; this alternative would measure harmful-completion detection on labeled outputs with false-positive monitoring; it does not establish attack success rate on adversarial web pages and zero unauthorized egress calls for the described use case.

D:

Isolate execution with restricted filesystem, network, credentials, resources, and observable policy denials; an instruction saying 'do not read secrets' cannot contain arbitrary code on an unrestricted host; this alternative would measure forbidden file and network access attempts blocked in execution tests; it does not establish attack success rate on adversarial web pages and zero unauthorized egress calls for the described use case.

E:

Layer input handling, model instructions, backend authorization, schema checks, and output review so one failed defense does not grant action authority; a single content filter cannot enforce account privileges or validate downstream business transactions; this alternative would measure denial of unauthorized actions and harmful outputs across multi-stage red-team paths; it does not establish attack success rate on adversarial web pages and zero unauthorized egress calls for the described use case.

Question 251

A regional risk review needs independent fraud, sanctions, and credit assessments before a final risk packet is assembled. What should the solution architect recommend?

  1. Version the handoff artifacts between agents
  2. Run independent specialists in parallel with a merge contract
  3. Delegate only specialized subtasks that add measurable value
  4. Bound handoffs, retries, and recursion across agents

Explanation

Correct Answer

B:

Parallel fan-out followed by deterministic aggregation can shorten wall-clock time if each branch has its own contract and errors are handled; this validates concurrency benefits without silently dropping a missing assessment; evaluate the design with p95 end-to-end latency and deterministic merge completeness under one-branch failures.

Incorrect Answers

A:

Immutable, versioned handoff artifacts and typed schemas prevent one agent from consuming ambiguous or mismatched intermediate state; artifact versioning governs exchange correctness, not whether the underlying market facts were researched accurately; this alternative would measure zero stale-artifact merges across concurrent revision and retry scenarios; it does not establish p95 end-to-end latency and deterministic merge completeness under one-branch failures for the described use case.

C:

Introduce a specialist only when its independent expertise, tool scope, or parallelism beats a simpler call on representative evaluations; more agents by themselves do not provide additional knowledge, authority, or better outcomes; this alternative would measure marginal quality improvement per added specialist against its incremental cost and latency; it does not establish p95 end-to-end latency and deterministic merge completeness under one-branch failures for the described use case.

D:

Enforce maximum depth, handoff count, timeouts, and explicit unresolved-case escalation in the orchestrator; recursive collaboration is not inherently self-terminating and cannot be made safe only by asking agents to cooperate; this alternative would measure zero unbounded cycles under adversarial unresolved tasks while retaining a useful escalation record; it does not establish p95 end-to-end latency and deterministic merge completeness under one-branch failures for the described use case.

Question 252

Claude generates fluent evidence summaries but cited links point only to a generic repository home page. Which intervention targets the defect rather than its symptoms?

  1. Evaluate retrieval quality independently of generation
  2. Preserve citation metadata through retrieval and synthesis
  3. Chunk documents along meaningful structural boundaries
  4. Filter retrieval by access control before model exposure
  5. Refresh the index when source content changes

Explanation

Correct Answer

B:

Store source IDs, locations, dates, and stable links with chunks and propagate them to answer citation validation; citation resolution testing verifies that evidence can actually be audited; verify the repair using fraction of material claims linked to resolvable source locations.

Incorrect Answers

A:

A stronger model cannot reliably cite evidence that retrieval never supplies; measure retrieval recall, ranking relevance, and evidence coverage on labeled query-document pairs before optimizing generation; that control is assessed by recall at k and evidence-coverage rate for known-answer queries; the incident instead requires fraction of material claims linked to resolvable source locations.

C:

Arbitrary equal-length chunks may split dependencies despite fitting an embedding model's input limit; chunk by headings and semantic sections with deliberate overlap where necessary while maintaining parent-document references; that control is assessed by retrieval of both governing clause and exception for labeled contract questions; the incident instead requires fraction of material claims linked to resolvable source locations.

D:

A model instruction to ignore confidential items is not a data-access boundary once the text is supplied; apply document permissions to retrieval candidates before material enters model context or logs; that control is assessed by unauthorized chunk exposure rate under staff permission matrices, targeting zero; the incident instead requires fraction of material claims linked to resolvable source locations.

E:

A larger context window cannot compensate for documents missing from the index; build a monitored ingestion path for adds, updates, deletes, and effective dates, with freshness checks and removal of superseded chunks; that control is assessed by publication-to-search freshness lag and absence of withdrawn clauses in results; the incident instead requires fraction of material claims linked to resolvable source locations.

Question 253

A lending team proposes customer profiling while source systems disagree on consent, fields, and retention rights. What is the most defensible next step?

  1. Map available data and its permitted uses before solution selection
  2. Convert vague accuracy demands into task-specific criteria
  3. Measure the current manual task before choosing automation
  4. Define which decisions remain with people
  5. Prioritize a repetitive step with bounded exceptions

Explanation

Correct Answer

A:

Data-readiness and usage-rights mapping identifies legitimate inputs, owners, quality gaps, and prohibited transfers before committing to an implementation; measured coverage of eligible sources exposes blockers that a high demo score conceals; evaluate the design with percentage of required fields with validated owners, quality checks, and approved processing purposes.

Incorrect Answers

B:

Convert vague requests into observable task outcomes, rubric-based quality dimensions, and failure thresholds that stakeholders can agree to test; a clear success rubric evaluates outcomes but cannot on its own implement retrieval access controls; this alternative would measure agreement between domain reviewers on a rubric and pass rate against a labeled acceptance set; it does not establish percentage of required fields with validated owners, quality checks, and approved processing purposes for the described use case.

C:

A baseline separates real business improvement from attractive demos; use completed cases, reviewer effort, errors, and exception handling as measurable comparators; process measurement establishes value, not the identity, data-permission, or runtime architecture of a Claude application; this alternative would measure change in verified minutes and rework per adjudicated claim versus the documented baseline; it does not establish percentage of required fields with validated owners, quality checks, and approved processing purposes for the described use case.

D:

Explicit decision-rights boundaries distinguish advisory generation from actions requiring a credentialed human or deterministic policy engine; decision ownership is an authorization and accountability control; it does not fix missing input records or prompt-token waste; this alternative would measure share of binding decisions showing the required named approver and rationale; it does not establish percentage of required fields with validated owners, quality checks, and approved processing purposes for the described use case.

E:

Select a high-frequency, well-described task with measurable work reduction and explicit escalation for ambiguous records; prioritization improves initial return and feasibility; it does not directly supply output-validation code or tenant isolation; this alternative would measure accepted cases per reviewer-hour with exception share tracked separately; it does not establish percentage of required fields with validated owners, quality checks, and approved processing purposes for the described use case.

Question 254

A regulated AI program has identified privacy and model-failure risks but no clear owners or residual-risk acceptance. What is the most targeted architecture choice?

  1. Record alternatives and rationale in an architecture decision record
  2. Keep a versioned decision and change history
  3. Maintain a risk register with controls and accountable owners
  4. Demonstrate known limitations alongside successful workflows

Explanation

Correct Answer

C:

Record probability and impact where appropriate, control owners, evidence, residual risk, and required approvals; a control register makes risk decisions reviewable rather than rhetorical; evaluate the design with completion and ownership of high-severity mitigations before the release gate.

Incorrect Answers

A:

Use an ADR recording context, considered options, decision, trade-offs, implications, and conditions for revisiting it; a final diagram alone often omits rejected alternatives and the reasoning behind the choice; this alternative would measure reconstructability of key design decisions after a simulated staff handoff; it does not establish completion and ownership of high-severity mitigations before the release gate for the described use case.

B:

Maintain a version-controlled decision log linked to implementation and review milestones, including superseded guidance; chat history and scattered slides rarely provide an authoritative current decision state; this alternative would measure correct retrieval of active decisions and their supersession chain during review; it does not establish completion and ownership of high-severity mitigations before the release gate for the described use case.

D:

Use representative happy-path and failure-path demonstrations with explicit fallback, uncertainty, and human ownership; a demo focused solely on best cases can mask residual risks and create unrealistic expectations; this alternative would measure stakeholder identification of boundaries and fallback responsibilities after the demo; it does not establish completion and ownership of high-severity mitigations before the release gate for the described use case.

Question 255

A helpdesk search pipeline uses a short, relevance-ranked evidence shortlist. Reviewers seed cases where the correct procedure is in the initial candidate pool but buried. Which result would justify the reranking step?

  1. Recall on human-authored paraphrases absent from manual headings
  2. Exact-ID hit rate at the top retrieval ranks
  3. Relevant-procedure rank and end-to-end answer support rate after reranking
  4. Conditional-rule accuracy with and without parent-document expansion

Explanation

Correct Answer

C:

Rank-sensitive evaluation tests whether useful evidence reaches the final context window; measure relevant-procedure rank and end-to-end answer support rate after reranking at the release gate; apply a query-aware reranker after broad retrieval, respecting freshness and permissions, to prioritize evidence for synthesis.

Incorrect Answers

A:

Paraphrase benchmark cases test the capability for which embeddings are chosen; vector retrieval can match conceptual similarity between user wording and document text, followed by evidence verification; this tests recall on human-authored paraphrases absent from manual headings; the actual release condition is relevant-procedure rank and end-to-end answer support rate after reranking.

B:

Identifier-focused tests reveal errors that a semantic relevance score may ignore; keyword or lexical matching is effective for codes, exact phrases, and terms whose characters carry the meaning; this tests exact-ID hit rate at the top retrieval ranks; the actual release condition is relevant-procedure rank and end-to-end answer support rate after reranking.

D:

Context reconstruction tests whether narrow hits are interpreted in their governing scope; use child-chunk retrieval for recall and fetch validated parent context or adjacent governing sections for interpretation; this tests conditional-rule accuracy with and without parent-document expansion; the actual release condition is relevant-procedure rank and end-to-end answer support rate after reranking.

Question 256

A finance analyst asks for quarterly totals and filtered aggregates across transactional records. What should the solution architect recommend?

  1. Query structured data through authorized SQL or typed APIs
  2. Use semantic vector search for paraphrased knowledge questions
  3. Retrieve parent context after finding a specific chunk
  4. Combine lexical and semantic candidates when queries mix both
  5. Rerank candidate passages for question relevance

Explanation

Correct Answer

A:

Use a governed structured query interface with parameter checks and business semantics for exact aggregates; a structured oracle detects missing rows and arithmetic mistakes; evaluate the design with exact aggregate agreement with a trusted query on authorized records.

Incorrect Answers

B:

Vector retrieval can match conceptual similarity between user wording and document text, followed by evidence verification; semantic search is not guaranteed to respect exact version strings, numerals, or access boundaries without additional controls; this alternative would measure recall on human-authored paraphrases absent from manual headings; it does not establish exact aggregate agreement with a trusted query on authorized records for the described use case.

C:

Use child-chunk retrieval for recall and fetch validated parent context or adjacent governing sections for interpretation; passing isolated small chunks may preserve relevance while losing the context needed for a correct conclusion; this alternative would measure conditional-rule accuracy with and without parent-document expansion; it does not establish exact aggregate agreement with a trusted query on authorized records for the described use case.

D:

Hybrid retrieval joins exact-term and semantic evidence, then applies metadata filtering and ranking; either modality alone leaves a known class of mixed-format queries uncovered; this alternative would measure joint exact-identifier accuracy and semantic-symptom recall on mixed queries; it does not establish exact aggregate agreement with a trusted query on authorized records for the described use case.

E:

Apply a query-aware reranker after broad retrieval, respecting freshness and permissions, to prioritize evidence for synthesis; a bigger first-stage candidate set without selection can dilute context and increase cost; this alternative would measure relevant-procedure rank and end-to-end answer support rate after reranking; it does not establish exact aggregate agreement with a trusted query on authorized records for the described use case.

Question 257

A billing-support pilot routes ordinary inquiries automatically and sends uncertain disputes to additional analysis or staff. Which evidence best shows that escalation catches costly mistakes without wasting capacity on every request?

  1. Macro-F1, critical-class recall, and cost per correctly triaged message on held-out traffic
  2. High-risk escalation recall and total cost under the required error ceiling
  3. Correct resolution of cross-exhibit conflicts at the required latency and review cost
  4. Correct equipment identification and recommendation accuracy on labeled photographs

Explanation

Correct Answer

B:

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

Incorrect Answers

A:

Joint quality and price measurement prevents choosing a cheap model that fails costly minority categories; compare an efficient Claude model against labeled classification cases and select the lowest-cost model that meets the defined accuracy floor; this tests macro-F1, critical-class recall, and cost per correctly triaged message on held-out traffic; the actual release condition is high-risk escalation recall and total cost under the required error ceiling.

C:

A targeted complex-case benchmark tests the capability that routine classification metrics miss; escalate tasks with demonstrated reasoning difficulty to a capable model after measuring gains on cases involving cross-document dependencies; this tests correct resolution of cross-exhibit conflicts at the required latency and review cost; the actual release condition is high-risk escalation recall and total cost under the required error ceiling.

D:

A multimodal acceptance set measures the capability absent from text-only trials; confirm that the chosen Claude interface supports the required input modality and test image quality, extraction, and reasoning together; this tests correct equipment identification and recommendation accuracy on labeled photographs; the actual release condition is high-risk escalation recall and total cost under the required error ceiling.

Question 258

An incident responder may ask Claude to inspect alerts but production firewall isolation must receive explicit authorization. Which strategy should be prioritized for this requirement?

  1. Enforce authorization inside the tool backend
  2. Record the user, tenant, and action in audit trails
  3. Verify tenant ownership of resource identifiers
  4. Require separate approval for high-impact mutations

Explanation

Correct Answer

D:

Bind high-impact actions to a separately verified approval and narrowly scoped execution credential; approval-path tests ensure the precise mutation was authorized before execution; evaluate the design with zero critical network changes without a recorded action-specific approval.

Incorrect Answers

A:

Implement enforceable role, amount, and resource checks at the action service, not just in natural-language instructions; prompt-only restrictions can be bypassed and are not an independently audited authorization control; this alternative would measure rejection of unauthorized tool calls even when sent outside the model workflow; it does not establish zero critical network changes without a recorded action-specific approval for the described use case.

B:

Propagate authenticated actor, tenant, delegated scope, request ID, and action outcome into tamper-resistant audit records; logging only a model response cannot attribute which authenticated principal caused a backend change; this alternative would measure audit reconstruction of actor, authority, target, and result for sampled mutations; it does not establish zero critical network changes without a recorded action-specific approval for the described use case.

C:

Check the requested resource against the authenticated tenant and role at the data service rather than trusting user-provided identifiers; syntactically valid IDs and a logged-in session do not establish ownership of a resource; this alternative would measure denial of cross-tenant identifier enumeration attempts in direct tool tests; it does not establish zero critical network changes without a recorded action-specific approval for the described use case.

Question 259

A billing assistant uses a browser and notification tool while handling confidential invoices. Which choice most directly satisfies the documented goal?

  1. Use claim verification to mitigate confident fabrication
  2. Threat-model prompt injection across lower-trust inputs
  3. Cap agent steps and resource expenditure
  4. Constrain outbound channels to prevent data exfiltration

Explanation

Correct Answer

D:

Restrict egress destinations, data classes, and tool arguments at the network and application layers, with audit trails; network- and tool-level evidence catches leakage not visible in user-facing text; evaluate the design with blocked unauthorized destinations and absence of confidential fields in egress logs.

Incorrect Answers

A:

Require support from retrieved source passages and provide an explicit unknown path for unsupported claims; fluency and citation formatting cannot establish factual support on their own; this alternative would measure proportion of numeric claims supported by independently inspected references; it does not establish blocked unauthorized destinations and absence of confidential fields in egress logs for the described use case.

B:

Model external content as untrusted input, test role spoofing and instruction attempts, and enforce downstream permissions regardless of the prompt; stronger wording in the system prompt is not a complete defense against an authorized dangerous tool; this alternative would measure unauthorized tool-action success rate over injected emails and documents; it does not establish blocked unauthorized destinations and absence of confidential fields in egress logs for the described use case.

C:

Set per-task step, token, cost, and elapsed-time budgets with a clear abort and escalation state; asking the model to 'be efficient' does not guarantee bounded autonomous execution; this alternative would measure maximum resource usage and orderly termination on deliberately unsolvable tasks; it does not establish blocked unauthorized destinations and absence of confidential fields in egress logs for the described use case.

Question 260

An enterprise platform wants automatic upgrades as soon as a newer Claude model identifier appears. What should the solution architect recommend?

  1. Gate model-version updates on regression tests
  2. Select a model path compatible with required modalities
  3. Use stronger reasoning capacity for genuinely complex decisions
  4. Evaluate model choice against the real input distribution

Explanation

Correct Answer

A:

Treat model replacements as versioned releases with task-specific eval gates, contract tests, canaries, and rollback capability; version-aware tests make improvement claims meaningful for the deployed application; evaluate the design with no contract regressions and acceptable quality/latency deltas on the proposed release.

Incorrect Answers

B:

Confirm that the chosen Claude interface supports the required input modality and test image quality, extraction, and reasoning together; model selection based solely on text benchmark accuracy does not prove image interpretation reliability; this alternative would measure correct equipment identification and recommendation accuracy on labeled photographs; it does not establish no contract regressions and acceptable quality/latency deltas on the proposed release for the described use case.

C:

Escalate tasks with demonstrated reasoning difficulty to a capable model after measuring gains on cases involving cross-document dependencies; high reasoning capacity is wasteful if the task only extracts a known field or matches a stable label; this alternative would measure correct resolution of cross-exhibit conflicts at the required latency and review cost; it does not establish no contract regressions and acceptable quality/latency deltas on the proposed release for the described use case.

D:

Representative workload evaluation must include input quality, document lengths, languages, and failure consequences before setting model policy; vendor benchmark rankings do not establish suitability for a domain-specific and skewed workload; this alternative would measure per-segment acceptance rate across language, length, and noisy-input strata; it does not establish no contract regressions and acceptable quality/latency deltas on the proposed release for the described use case.

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