Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 18

Question 170

An MCP service encounters errors while fetching customer records. A requested account number is well formed but does not exist. What should the architect specify in the design?

  1. Return a structured rate-limit error and retry-after guidance
  2. Return a structured validation error identifying the affected field
  3. Return a permission-denied outcome and require an authorized path
  4. Classify the issue as transient unavailable and provide a safe diagnostic identifier
  5. Return a distinct not-found status without recommending blind retries

Explanation

Correct Answer

E:

A genuinely missing resource is different from transient infrastructure failure; retrying unchanged identifiers adds load without resolving it.

Incorrect Answers

A:

This measure fits a different trigger: a backend returns HTTP 429 with a retry window. In that setting, Throttling is transient and should be handled with bounded delay, not treated as an absent customer record; it does not resolve the present need to report a terminal not-found outcome.

B:

A field-specific error supports a corrected retry, whereas generic failure text makes it hard to choose an appropriate next action; it protects against the case where an invalid date format fails argument validation before tool execution, rather than the failure described here: a requested account number is well formed but does not exist.

C:

A permission error cannot be fixed by repeating the same call; it needs identity correction or escalation without leaking account details; that applies when a tool call is rejected because the caller lacks access to the account, whereas this case requires a response to a requested account number is well formed but does not exist.

D:

Clear error categorization lets orchestration retry selectively while keeping internal stack traces out of model-facing content; this is useful for a database loses connectivity during a lookup, but it leaves the stated requirement to report a terminal not-found outcome unmet.

Question 171

A research assistant uses the Messages API to coordinate evidence gathering. An assistant text block describes a search but includes no tool_use block. Which change most directly addresses this condition?

  1. Detect max_tokens and continue or retry within a controlled budget
  2. Treat tool output as data and enforce higher-priority instructions and tool permissions
  3. Wait for the first result and derive the second request from verified output
  4. Treat ordinary assistant text as content rather than an executable tool instruction

Explanation

Correct Answer

D:

Only explicit tool-use content requests host execution; natural-language statements do not authorize a network call.

Incorrect Answers

A:

This measure fits a different trigger: a response ends because the output-token limit was reached halfway through a structured answer. In that setting, The stop reason distinguishes token exhaustion from a finished response; blindly accepting the fragment risks malformed or unsupported findings; it does not resolve the present need to avoid performing unrequested external work.

B:

External pages may contain prompt injection; tool output should inform answers, not override policy or initiate new privileges; that applies when a search tool returns untrusted web text that says to ignore system instructions, whereas this case requires a response to an assistant text block describes a search but includes no tool_use block.

C:

When the next arguments depend on earlier evidence, speculative parallel execution can issue an invalid or misleading query; it protects against the case where two sequential searches must use the first response to form the second query, rather than the failure described here: an assistant text block describes a search but includes no tool_use block.

Question 172

A product team decomposes complex Claude tasks into subtasks. A report must combine three unrelated data sources before drafting. What is the most defensible next step?

  1. Call Claude once with a precise classification instruction and output contract
  2. Implement a deterministic pipeline around Claude calls rather than a free-form planner
  3. Extract, validate, calculate, then explain in that order
  4. Collect sources concurrently and synthesize once all results pass validation

Explanation

Correct Answer

D:

Independent lookups can run together, but writing should wait for the join to prevent omissions. This matches the workflow goal to parallelize independent evidence acquisition rather than introducing an unrelated action.

Incorrect Answers

A:

This measure fits a different trigger: a customer email requires one classification label and no external data. In that setting, A single well-scoped task avoids unnecessary agent coordination when no independent dependencies or tools are needed; it does not resolve the present need to parallelize independent evidence acquisition.

B:

Known stable stages are easier to test and audit as code; open-ended planning adds variance without meeting an unmet requirement; it protects against the case where a workflow repeats the same fixed five stages for every case, rather than the failure described here: a report must combine three unrelated data sources before drafting.

C:

The arithmetic is only meaningful after source values pass type and completeness checks; dependencies determine stage order; this is useful for an analysis must validate extracted figures before calculating a ratio, but it leaves the stated requirement to parallelize independent evidence acquisition unmet.

Question 173

A company processes thousands of independent documents through Claude. A support agent needs a reply within the current live conversation. Which approach satisfies the stated constraint?

  1. Attach stable custom identifiers and reconcile outputs with the submission manifest
  2. Record per-request success or error and retry only eligible failures
  3. Use a normal request-response call for the interactive decision
  4. Record the document revision or digest with each submitted request

Explanation

Correct Answer

C:

Deferred batch completion is a poor fit when a customer is waiting for an immediate answer; workload latency requirements control the choice.

Incorrect Answers

A:

Result ordering alone may not match input ordering; explicit IDs prevent extracted data being assigned to the wrong document; this is useful for thousands of batch results must be matched back to source documents, but it leaves the stated requirement to choose synchronous processing for latency-sensitive work unmet.

B:

A batch is not necessarily all-or-nothing; item-level bookkeeping avoids rerunning already successful documents; that applies when one batch item fails while the remaining documents complete, whereas this case requires a response to a support agent needs a reply within the current live conversation.

D:

Versioned input identity ensures results are interpreted against the data actually sent, not a later file with different contents; it protects against the case where the source file changes after the batch is submitted, rather than the failure described here: a support agent needs a reply within the current live conversation.

Question 174

A team maintains path-aware architecture conventions in Claude Code. Only infrastructure-as-code files require a plan-before-change policy. Which mechanism should be selected?

  1. Revise path patterns and test their effective selection
  2. Keep shared guidance in one place and scoped exceptions explicitly narrow
  3. Constrain edit permissions to approved paths at the execution layer
  4. Attach the review procedure to infrastructure configuration patterns

Explanation

Correct Answer

D:

A path-aware rule gives extra caution where necessary without slowing unrelated text edits. The relevant requirement is to scope infrastructure planning requirements to matching files, and this measure operates at that decision point.

Incorrect Answers

A:

Rule files are not intelligent about renamed folders; matching must be maintained alongside structural refactors; this is useful for a new package directory does not match the existing rules, but it leaves the stated requirement to scope infrastructure planning requirements to matching files unmet.

B:

Redundant copies can drift; exceptions should state exactly what changes for a subset of files; that applies when a scoped rule repeats a contradictory general rule verbatim, whereas this case requires a response to only infrastructure-as-code files require a plan-before-change policy.

C:

Scoped instruction text improves intent but does not prevent a capable tool from writing elsewhere when misused; it protects against the case where a critical write restriction is written only as path-specific prose, rather than the failure described here: only infrastructure-as-code files require a plan-before-change policy.

Question 175

An agent controller must coordinate recoverable workflows across interruptions. An interrupted research task has several finished specialist outputs. What should be implemented at the appropriate boundary?

  1. Load checkpoint status and reuse completed result artifacts
  2. Reauthorize actions against the current permission configuration
  3. Create a clean conversation and explicitly scoped context for the new task
  4. Persist origin session, fork lineage, task ID, and artifact references
  5. Replay the persisted, ID-correlated tool_result into the restored conversation

Explanation

Correct Answer

A:

A checkpointed task graph supports partial recovery and avoids paying for identical completed work. This matches the workflow goal to resume only unfinished work rather than introducing an unrelated action.

Incorrect Answers

B:

Session continuity is not continuing authorization; permissions may change independently of the saved transcript; this is useful for a worker changed permissions in an earlier session, but today's role policy differs, but it leaves the stated requirement to resume only unfinished work unmet.

C:

Starting fresh avoids accidentally promoting data from an unrelated case into a new workflow; that applies when a new agent task accidentally inherits old customer contact details, whereas this case requires a response to an interrupted research task has several finished specialist outputs.

D:

Provenance across resumes and forks is essential for diagnosing which state or branch produced a disputed conclusion; it protects against the case where engineers need to trace exactly where a resumed result originated, rather than the failure described here: an interrupted research task has several finished specialist outputs.

E:

This measure fits a different trigger: a tool result was durably saved with its request ID, but the host crashed before passing it back to Claude. In that setting, When the result is already durably recorded, the host can resume the exact tool-use/result exchange without repeating the side-effecting operation; it does not resolve the present need to resume only unfinished work.

Question 176

A support application selects tool-use policies for sensitive operations. A tool can issue refunds but the current conversation has not passed identity verification. Which option best meets the operational need?

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

Explanation

Correct Answer

C:

Tool choice is not the same as permission; the financial action must be gated on verified identity and policy. For the stated case, the check is whether the system can deny the mutation until verification is complete; this approach supplies that control.

Incorrect Answers

A:

Retiring ambiguous or obsolete tools attacks the selection problem at its source instead of adding reminders to avoid the old one; it protects against the case where an agent repeatedly chooses an obsolete lookup despite a new supported endpoint, rather than the failure described here: a tool can issue refunds but the current conversation has not passed identity verification.

B:

Over-forcing one retrieval tool can cause needless calls; controlled automatic choice suits variable informational needs; that applies when a user asks an open-ended policy question that does not necessarily require a tool, whereas this case requires a response to a tool can issue refunds but the current conversation has not passed identity verification.

D:

The none mode prevents tool calls on that request even when tool definitions are present; later actions still need application-side authorization and controls; this is useful for an audit-only response must not invoke any available tools while explaining a policy decision, but it leaves the stated requirement to deny the mutation until verification is complete unmet.

Question 177

A document-processing agent decides whether to retry failed MCP calls. A tool reports that the user's approval is missing for a payout. What is the most appropriate design choice?

  1. Use an operation ID to query authoritative completion status
  2. Return a field-level constraint violation for the caller to repair
  3. Treat missing approval as a policy precondition requiring escalation
  4. Return a password-protection status rather than a fabricated empty conversion

Explanation

Correct Answer

C:

Approval cannot be manufactured by model retries; the workflow must obtain an authorized decision outside the model.

Incorrect Answers

A:

Retrying a financial mutation blindly can duplicate a transaction; checking durable state is safer than inferring failure from a timeout; that applies when a network timeout leaves the outcome of a state-changing operation uncertain, whereas this case requires a response to a tool reports that the user's approval is missing for a payout.

B:

A type or range error requires argument correction, not infrastructure backoff or an alternate source search; it protects against the case where a validator rejects a confidence value outside its supported range, rather than the failure described here: a tool reports that the user's approval is missing for a payout.

D:

A clear input limitation enables the user to provide an accessible document through approved channels; repeated conversion is ineffective; this is useful for an XML converter receives a password-protected file it cannot read, but it leaves the stated requirement to route to human authorization instead of looping unmet.

Question 178

An internal compliance agent handles ambiguous requests. Two approved policy documents contain contradictory effective dates. Which response is most reliable?

  1. Seek the necessary proof before making the decision
  2. Request compliance-owner guidance instead of inferring a rule
  3. Record both versions and request adjudication from the policy owner
  4. Record the verified conditions and perform the permitted action

Explanation

Correct Answer

C:

Silent selection of whichever date looks newest can be wrong when different scopes or transition rules apply. Here it serves the explicit need to surface the conflict for resolution without adding a separate unverified inference.

Incorrect Answers

A:

Eligibility text is not enough when required supporting facts are absent; clarification can close the gap without inventing evidence; that applies when a document explicitly allows an action, but the request lacks required evidence, whereas this case requires a response to two approved policy documents contain contradictory effective dates.

B:

This measure fits a different trigger: a new regulation is mentioned but the maintained policy library has not been updated. In that setting, Regulatory ambiguity cannot be repaired by confident paraphrase; the accountable team must supply the applicable policy; it does not resolve the present need to surface the conflict for resolution.

D:

Reliable systems should distinguish genuine blockers from resolved ones and avoid unnecessary repeated escalation; it protects against the case where a reviewer confirms that all documented conditions are satisfied, rather than the failure described here: two approved policy documents contain contradictory effective dates.

Question 179

A developer asks Claude Code to implement a new serialization routine. The desired behavior is defined by a small set of stable examples. What should the team do to resolve this particular gap?

  1. Write or update tests that encode the expected outputs before implementation
  2. Ask for representative input/output pairs and edge cases
  3. Inspect complexity assumptions and test evidence against acceptance criteria
  4. Use the failing test output to make a targeted correction

Explanation

Correct Answer

A:

A test-first loop makes success measurable and gives the agent rapid feedback about whether its code matches the spec.

Incorrect Answers

B:

This measure fits a different trigger: the exact expected encoding is ambiguous in the feature request. In that setting, Concrete examples expose ambiguity in the contract before implementation, reducing incorrect assumptions that tests might otherwise miss; it does not resolve the present need to turn the examples into executable tests.

C:

A concise rationale should be backed by measured or inspectable behavior, not by the model's own confidence; it protects against the case where a developer wants to know why the proposed algorithm is acceptable, rather than the failure described here: the desired behavior is defined by a small set of stable examples.

D:

Compilation checks syntax and types, not all behavior; a failing edge-case test provides precise evidence for improvement; this is useful for Claude produces a patch that compiles but breaks a boundary case, but it leaves the stated requirement to turn the examples into executable tests unmet.

Question 180

A research pipeline asks Claude to emit machine-readable evidence records. The model invents unrecognized status strings such as 'sort-of-verified'. Which control is most directly relevant?

  1. Use a list of claim objects, each with its own source references and stance
  2. Check source IDs against the approved evidence registry after parsing
  3. Use an enum for allowed verification statuses
  4. Allow a conditional locator such as a page or URL fragment as appropriate
  5. Define source_id and publication_date as required schema properties

Explanation

Correct Answer

C:

Enumerated values make downstream control flow predictable and prevent speculative status text from being treated as a recognized gate.

Incorrect Answers

A:

Structured claim-level provenance prevents competing interpretations from being collapsed into one misleading free-text conclusion; that applies when results need to preserve multiple contradictory claims from one source set, whereas this case requires a response to the model invents unrecognized status strings such as 'sort-of-verified'.

B:

JSON Schema checks shape and types, not whether a purported reference actually exists or supports the claim; it protects against the case where a tool-call payload satisfies syntax but cites an unknown document ID, rather than the failure described here: the model invents unrecognized status strings such as 'sort-of-verified'.

D:

The schema should reflect source-type differences; forcing a PDF page number for HTML would create meaningless metadata; this is useful for some sources have no stable page number because they are HTML, but it leaves the stated requirement to restrict outputs to approved states unmet.

E:

This measure fits a different trigger: a citation must include both source ID and publication date. In that setting, Required fields expose missing citation metadata early instead of allowing evidence records that cannot later be traced; it does not resolve the present need to restrict outputs to approved states.

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