Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 34

Question 330

A support application selects tool-use policies for sensitive operations. A user asks an open-ended policy question that does not necessarily require a tool. Which mechanism should be selected?

  1. Update the tool registry and descriptions so only the current lookup is eligible
  2. Permit automatic tool selection within a safe allowlist
  3. Expose the scoped order-status read tool to the support agent
  4. Keep the refund capability unavailable or execution-blocked for this state
  5. Set tool_choice to none on the request that should not dispatch a tool

Explanation

Correct Answer

B:

Over-forcing one retrieval tool can cause needless calls; controlled automatic choice suits variable informational needs.

Incorrect Answers

A:

Retiring ambiguous or obsolete tools attacks the selection problem at its source instead of adding reminders to avoid the old one; the alternative would make sense if an agent repeatedly chooses an obsolete lookup despite a new supported endpoint. It does not directly fix the observed condition: a user asks an open-ended policy question that does not necessarily require a tool.

C:

For the separate situation where a customer asks for current order status, and the service has a read-only order API, a narrow read capability can answer the question while preserving the boundary against unauthorized refunds or edits; the current task is to allow the model to answer or fetch as needed.

D:

This control matters when a tool can issue refunds but the current conversation has not passed identity verification; tool choice is not the same as permission; the financial action must be gated on verified identity and policy. The required decision is different: a user asks an open-ended policy question that does not necessarily require a tool.

E:

The none mode prevents tool calls on that request even when tool definitions are present; later actions still need application-side authorization and controls; that assumes an audit-only response must not invoke any available tools while explaining a policy decision. The item tests the different ability to allow the model to answer or fetch as needed.

Question 331

An AI team uses a generator and reviewer for compliance reports. The reviewer identifies one unsupported paragraph among otherwise valid sections. What should be implemented at the appropriate boundary?

  1. Provide the reviewer the cited sources and a claim-verification rubric
  2. Report the claim, evidence gap, and suggested bounded repair
  3. Preserve both positions and send the evidence to a human approver
  4. Keep review decisions in a distinct controlled stage or human gate

Explanation

Correct Answer

B:

Targeted feedback lets the writer fix the failing span without regenerating a document that already passed most checks.

Incorrect Answers

A:

This control matters when the reviewer has no access to source documents and simply judges fluency; without reference evidence, the reviewer can only assess plausibility and may approve fabricated claims. The required decision is different: the reviewer identifies one unsupported paragraph among otherwise valid sections.

C:

Persistent disagreement is a signal for accountable judgment, not a reason to keep agents debating indefinitely; the alternative would make sense if a dispute remains after the independent reviewer raises a high-risk concern. It does not directly fix the observed condition: the reviewer identifies one unsupported paragraph among otherwise valid sections.

D:

An independent check loses value if the generator can overwrite or ignore a failed result without accountability; that assumes the writer is allowed to alter the reviewer verdict before publication. The item tests the different ability to return actionable localized corrections.

Question 332

A support workflow handles errors from external services. A backend returns a permanent account-permission denial. Which option best meets the operational need?

  1. Treat a validated empty optional field as data, not a tool outage
  2. Query the transaction ledger by idempotency key before retrying
  3. Convert the exception to a safe user-facing status and retain trace IDs internally
  4. Stop the action and route to the approved access process

Explanation

Correct Answer

D:

A fixed permission constraint is not repaired by exponential backoff; repeated calls can add noise without changing authority.

Incorrect Answers

A:

An expected empty value is distinct from execution failure; confusing them causes unnecessary recovery attempts; that assumes a policy lookup returns a valid empty result for an optional note. The item tests the different ability to avoid transient retry logic for authorization errors.

B:

This control matters when a refund call may have succeeded before its network connection dropped; unknown delivery of a response is not proof the operation failed; reconciliation prevents duplicate financial effects. The required decision is different: a backend returns a permanent account-permission denial.

C:

The user needs an accurate next step, not sensitive implementation paths or credentials embedded in a service error; the alternative would make sense if a customer-facing workflow receives a detailed internal stack trace. It does not directly fix the observed condition: a backend returns a permanent account-permission denial.

Question 333

A research MCP server uses structured errors and pagination. Only one of five federated indexes fails while others succeed. What is the most appropriate design choice?

  1. Return successful index results plus per-source error statuses
  2. Return an unsupported-format status with the detected media type
  3. Expose a sanitized error code and trace ID while logging restricted details internally
  4. Return a distinct expired-cursor error with a restart strategy

Explanation

Correct Answer

A:

A mixed outcome should retain usable evidence while informing the coordinator that coverage is incomplete. The relevant requirement is to report partial success explicitly, and this measure operates at that decision point.

Incorrect Answers

B:

A format mismatch needs a different parser or workflow, not repeated identical calls to an unavailable feature; that assumes an article is accessible but its content type is not supported by the parser. The item tests the different ability to report partial success explicitly.

C:

Structured diagnostics allow troubleshooting without dumping sensitive service internals into a language-model conversation; the alternative would make sense if a request fails with an internal stack trace containing server paths. It does not directly fix the observed condition: only one of five federated indexes fails while others succeed.

D:

For the separate situation where a returned cursor has expired between paginated requests, an invalid cursor is not 'no more results'; preserving the error prevents a partial search from appearing complete; the current task is to report partial success explicitly.

Question 334

A Claude Agent SDK workflow delegates repository investigations to subagents. A parent agent needs a specialist to take a bounded action rather than just offer advice. Which response is most reliable?

  1. Have the worker report changed or inspected paths and evidence for its conclusion
  2. Run the investigation in a subagent and return a concise evidence-based summary
  3. Launch the independent module inquiries in parallel and merge results later
  4. Give the Task-oriented specialist a concrete goal, scope, and expected handoff

Explanation

Correct Answer

D:

Structured delegated tasks clarify which worker owns the operation and how its outcome will return to the parent. The relevant requirement is to delegate an explicit unit of work through the supported agent task mechanism, and this measure operates at that decision point.

Incorrect Answers

A:

The parent can validate a specialist's claims only when the output identifies what was examined or modified; the alternative would make sense if a subagent returns a broad recommendation without citing inspected files. It does not directly fix the observed condition: a parent agent needs a specialist to take a bounded action rather than just offer advice.

B:

This control matters when a specialist's large file-search transcript is not useful to the parent; an isolated worker context avoids flooding the coordinator with irrelevant file reads while preserving the actionable findings. The required decision is different: a parent agent needs a specialist to take a bounded action rather than just offer advice.

C:

Concurrent subagents can reduce wall-clock time when their files and decisions do not have data dependencies; that assumes two independent test modules can be investigated separately. The item tests the different ability to delegate an explicit unit of work through the supported agent task mechanism.

Question 335

A coding assistant reviews generated patch descriptions for consistency. The team's examples omit multi-file changes. What should the team do to resolve this particular gap?

  1. Revise labels and re-evaluate examples against the active review criteria
  2. Use clearly illustrative placeholders and require current identifiers from the input
  3. Provide few-shot examples that distinguish behavior-preserving and behavior-changing patches
  4. Annotate samples with the behavior or contract difference that changes the classification
  5. Add a small labeled example covering coordinated changes across multiple modules

Explanation

Correct Answer

E:

A representative complex case teaches how to summarize a unit of work that spans files without treating each edit as unrelated.

Incorrect Answers

A:

Examples are policy-bearing inputs; inconsistency between versions can undermine otherwise precise instructions; the alternative would make sense if an example is accurate for old review policy but not the current standard. It does not directly fix the observed condition: the team's examples omit multi-file changes.

B:

Few-shot examples guide format but must not become a source of fabricated ticket IDs or commit hashes; that assumes the assistant copies placeholder identifiers from example output. The item tests the different ability to represent composite patch cases.

C:

For the separate situation where some changes are wrongly described as refactors rather than bug fixes, contrastive labels teach the category distinction that ambiguous prose alone may not reliably convey; the current task is to represent composite patch cases.

D:

This control matters when a reviewer cannot tell why two visually similar changes receive different labels; the important signal is semantic change, not file size or formatting; rationale-bearing examples make that visible. The required decision is different: the team's examples omit multi-file changes.

Question 336

A customer-support assistant must decide when to request human assistance. A customer reports urgent harm that the assistant cannot safely assess. Which control is most directly relevant?

  1. Pause account changes and request approved verification or support review
  2. Apply the verified rule within delegated permission and record the decision
  3. Create a human-approval request with the verified amount and case evidence
  4. Trigger the organization's urgent human handoff process

Explanation

Correct Answer

D:

High-impact, ambiguous reports warrant a trained responder and documented procedure rather than an autonomous improvised resolution.

Incorrect Answers

A:

This control matters when the user provides conflicting account ownership details; conflicting identity evidence makes a mutation unsafe; a fluent guessed choice is not a substitute for authentication. The required decision is different: a customer reports urgent harm that the assistant cannot safely assess.

B:

Human review should be focused on risk or uncertainty; routine well-defined cases can proceed under controlled authorization; the alternative would make sense if a support specialist finds the policy clear and customer eligibility already verified. It does not directly fix the observed condition: a customer reports urgent harm that the assistant cannot safely assess.

C:

Authority limits must be observed at execution; the model cannot increase its own financial permissions; that assumes a refund would exceed the support agent's approved limit. The item tests the different ability to use a defined high-priority escalation path.

Question 337

A multi-agent developer assistant manages long-running refactoring tasks. A worker wants to reuse another agent's whole transcript as context. How should the workflow be adjusted?

  1. Mark stale notes superseded and record the new source revision
  2. Keep a checklist of prerequisites and blockers in the compacted handoff
  3. Require entries keyed by task, file path, and evidence status
  4. Pass a summary with linked files, decisions, and unresolved questions

Explanation

Correct Answer

D:

Raw transcripts contain irrelevant exploration and private context; scoped artifacts support collaboration more efficiently.

Incorrect Answers

A:

Scratchpads are working memory, not authority; current repository evidence must win over obsolete summaries; the alternative would make sense if an old scratchpad conflicts with an authoritative code search performed today. It does not directly fix the observed condition: a worker wants to reuse another agent's whole transcript as context.

B:

Losing a prerequisite can cause later stages to execute prematurely even when the prose summary appears coherent; that assumes a compaction step drops the list of migration prerequisites. The item tests the different ability to share distilled artifacts rather than entire conversations.

C:

For the separate situation where multiple workers create unrelated notes with no common file identifiers, a consistent scratchpad makes cross-worker discoveries retrievable and reduces conflicting or orphaned notes; the current task is to share distilled artifacts rather than entire conversations.

Question 338

A team wants repeatable Claude Code workflows for contributors. A command needs a target module supplied by the developer. Which implementation decision is the best fit?

  1. Design the command to accept and validate a module argument
  2. Package the instructions and supporting resources as a Claude Code Skill
  3. Keep a concise skill overview and reference supplemental files as needed
  4. Create a project-scoped custom command with clear inputs and instructions

Explanation

Correct Answer

A:

Parameterized workflows remain reusable while tying file operations to a caller-specified scope instead of assuming a default.

Incorrect Answers

B:

This control matters when a repository has a capability that Claude should discover when task context matches; skills can encapsulate specialized procedures with descriptions that support appropriate discovery and reuse. The required decision is different: a command needs a target module supplied by the developer.

C:

Progressive disclosure preserves context while still allowing detailed procedures to be consulted for the actual task; the alternative would make sense if a skill includes several bulky examples rarely needed for selection. It does not directly fix the observed condition: a command needs a target module supplied by the developer.

D:

For the separate situation where developers repeatedly invoke a fixed changelog-generation routine by name, a named command makes an explicit routine easy to invoke consistently without embedding the same instructions in every chat; the current task is to make the invocation parameter explicit.

Question 339

A Claude Code repository is organized into multiple services with different local rules. The team's root file contains a temporary incident note from last month. What should the architect specify in the design?

  1. Use actual tool permissions and approval workflows for risky commands
  2. Update the path or import so Claude Code can load the maintained guidance
  3. Use a supported CLAUDE.md import or concise reference to the maintained document
  4. Delete or archive obsolete operational details from shared instructions

Explanation

Correct Answer

D:

Outdated incident context can mislead future tasks; persistent memory should contain durable conventions and verified references.

Incorrect Answers

A:

Instruction files can steer behavior but do not grant or enforce operating-system authorization by themselves; the alternative would make sense if a developer assumes an automatically loaded file authorizes privileged commands. It does not directly fix the observed condition: the team's root file contains a temporary incident note from last month.

B:

A prompt reminder cannot replace a broken import; resolving the file reference restores the policy's discoverability; that assumes an imported policy file is no longer present after a directory move. The item tests the different ability to remove time-limited noise from persistent memory.

C:

This control matters when a release policy document already exists and should not be copied into multiple instruction files; importing maintained guidance avoids parallel copies that drift when the release process changes. The required decision is different: the team's root file contains a temporary incident note from last month.

Question 340

A document-extraction service must report the state of every processed page. A conversion worker crashes but a coordinator has cached its last successful output from another document. Which change most directly addresses this condition?

  1. Return valid page results and flag page nine for targeted correction
  2. Report passed, failed, and needs-review counts separately
  3. Return an explicit missing-required-field failure
  4. Use document identity to prevent reuse of unrelated cached outputs
  5. Return expected and observed totals with a validation-failure code

Explanation

Correct Answer

D:

Cached success for a different input is not a valid substitute; identity-bound artifacts prevent accidental data contamination.

Incorrect Answers

A:

For the separate situation where eight pages are parsed, but page nine contains unreadable characters, page-level status avoids losing successful extraction while preventing the unreadable page from being misrepresented as verified; the current task is to reject stale cross-document fallback.

B:

Accurate status aggregation ensures that human-review items remain visible instead of being included in successful automation totals; the alternative would make sense if a final report says all invoices passed even though a few were escalated. It does not directly fix the observed condition: a conversion worker crashes but a coordinator has cached its last successful output from another document.

C:

This control matters when a required identity field is missing from the extracted record; a ledger-ready record must satisfy required business fields; a syntactically partial result cannot silently pass to commit. The required decision is different: a conversion worker crashes but a coordinator has cached its last successful output from another document.

E:

Specific diagnostics help corrective steps distinguish arithmetic conflict from parser failure or missing pages; that assumes a semantic validator detects inconsistent tax arithmetic. The item tests the different ability to reject stale cross-document fallback.

Claude Certified Architect - Foundations (CCAR-F) (CCA-F) Practice Tests