Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 39

Question 380

A document-processing agent decides whether to retry failed MCP calls. A network timeout leaves the outcome of a state-changing operation uncertain. What is the most defensible next step?

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

Explanation

Correct Answer

C:

Retrying a financial mutation blindly can duplicate a transaction; checking durable state is safer than inferring failure from a timeout.

Incorrect Answers

A:

A type or range error requires argument correction, not infrastructure backoff or an alternate source search; the alternative would make sense if a validator rejects a confidence value outside its supported range. It does not directly fix the observed condition: a network timeout leaves the outcome of a state-changing operation uncertain.

B:

A clear input limitation enables the user to provide an accessible document through approved channels; repeated conversion is ineffective; that assumes an XML converter receives a password-protected file it cannot read. The item tests the different ability to reconcile mutation state before retrying.

D:

This control matters when a tool reports that the user's approval is missing for a payout; approval cannot be manufactured by model retries; the workflow must obtain an authorized decision outside the model. The required decision is different: a network timeout leaves the outcome of a state-changing operation uncertain.

E:

For the separate situation where a service responds with a temporary capacity error and a recommended wait, temporary saturation may clear; paced retries can recover without causing a synchronized load spike; the current task is to reconcile mutation state before retrying.

Question 381

A customer-service assistant may hand tasks to billing, returns, and policy specialists. A returns worker needs a decision made by the identity-verification worker. Which approach satisfies the stated constraint?

  1. Make identity verification an explicit prerequisite in the coordinator workflow
  2. Check the current case state and ownership before starting an escalated job
  3. Configure only the needed read-only policy tools for that specialist
  4. Pass narrowly scoped task context and redact irrelevant customer information

Explanation

Correct Answer

A:

A worker should not process a customer-specific return against an unverified identity; graph ordering enforces the prerequisite.

Incorrect Answers

B:

State validation at handoff keeps asynchronous specialists from acting on superseded requests or producing redundant outcomes; the alternative would make sense if an escalation specialist is assigned a case that has already been resolved. It does not directly fix the observed condition: a returns worker needs a decision made by the identity-verification worker.

C:

Least-privilege tool access prevents a policy task from accidentally or maliciously exercising financial operations; that assumes a policy specialist can answer without any payment-system access. The item tests the different ability to block dependent work until verification completes.

D:

For the separate situation where each specialist should see only the customer details it needs, a smaller context packet reduces leakage and helps a specialist focus on the authorized decision instead of broad account history; the current task is to block dependent work until verification completes.

Question 382

A multi-agent research controller launches short-lived specialists. The coordinator cannot tell whether a worker finished or merely started searching. Which mechanism should be selected?

  1. Require status, result, and unresolved-work fields in each worker handoff
  2. Assign each worker a purpose-specific allowed-tool set
  3. Set agent-level execution limits and monitor per-task consumption
  4. Pass a curated task brief and the approved evidence only

Explanation

Correct Answer

A:

A typed completion status allows the parent to distinguish finished evidence collection from a partial exploratory result.

Incorrect Answers

B:

For the separate situation where a finance analyst must not obtain the searcher's CRM write tool, tool permissions should follow delegated duties, not the maximum privileges of the parent or another worker; the current task is to return explicit completion state from specialists.

C:

Budget limits stop a narrow delegated task from silently exhausting the resources allocated to the entire orchestration; that assumes a worker is likely to call an expensive external tool excessively. The item tests the different ability to return explicit completion state from specialists.

D:

This control matters when a writing specialist receives the entire private search history, including unrelated leads; specialized workers benefit from an explicit information boundary; inherited unrelated history can confuse task scope and expose sensitive data. The required decision is different: the coordinator cannot tell whether a worker finished or merely started searching.

Question 383

An enterprise design team compares agent organization patterns for different requests. The same specialized translation step repeats across many workflows. What should be implemented at the appropriate boundary?

  1. Delegate independent evidence collection then combine results in a coordinator
  2. Expose a narrowly defined reusable translation worker behind a clear input/output contract
  3. Use sequential stages with explicit prerequisite checks
  4. Use one constrained model call instead of a multi-agent chain

Explanation

Correct Answer

B:

A well-defined specialist reduces duplicated orchestration logic while keeping translation behavior consistent across callers.

Incorrect Answers

A:

This control matters when an investigation has three independent evidence streams and one final answer; this topology exploits concurrency without letting each worker independently publish a different final answer. The required decision is different: the same specialized translation step repeats across many workflows.

C:

The second stage depends on verified clauses, so parallel dispatch would let it act on absent or provisional data; that assumes a legal review cannot begin until a contract-clause extractor finishes. The item tests the different ability to reuse a stable specialist boundary.

D:

For the separate situation where a small deterministic classification needs no external calls, simple classification has no meaningful subtask parallelism; multiple agents add latency and reconciliation cost without improving the required outcome; the current task is to reuse a stable specialist boundary.

Question 384

A support-orchestration service decides how to pass tasks between specialists. A delegated task completes but its result schema has an unknown status. Which option best meets the operational need?

  1. Select a named specialist based on intent and permission boundaries
  2. Treat the unrecognized status as invalid and request clarification or manual review
  3. Associate delegated jobs with stable identifiers and recoverable progress
  4. Validate required fields before invoking the specialist

Explanation

Correct Answer

B:

A closed status enumeration prevents unfamiliar values from accidentally being treated as success. The relevant requirement is to fail safely at the handoff boundary, and this measure operates at that decision point.

Incorrect Answers

A:

For the separate situation where a customer intent is classified, but no worker has been chosen, an explicit dispatch decision links a validated customer need to a worker with relevant expertise and tool access; the current task is to fail safely at the handoff boundary.

C:

A durable job identity connects resumes to prior work and avoids launching duplicate specialists blindly; the alternative would make sense if the coordinator wants to know whether a stalled task can be resumed. It does not directly fix the observed condition: a delegated task completes but its result schema has an unknown status.

D:

Input checks at the coordinator boundary reduce wasted agent turns and expose missing fields early; that assumes a specialist rejects an input because it lacks a case identifier. The item tests the different ability to fail safely at the handoff boundary.

Question 385

A research publishing workflow uses confidence-aware human oversight. An automated report cites a source that cannot be retrieved by the reviewer. What is the most appropriate design choice?

  1. Require a verifiable source or explicit review of the unsupported claim
  2. Analyze error patterns and update targeted tests or escalation criteria
  3. Flag both claims and their evidence for accountable adjudication
  4. Apply the checked fix without adding a human approval queue
  5. Use observed error rates or rubric-based review rather than the unsupported percentage

Explanation

Correct Answer

A:

Publication quality depends on accessible supporting material, not merely on a plausible citation string. The relevant requirement is to withhold high-impact release pending evidence, and this measure operates at that decision point.

Incorrect Answers

B:

Systematic override patterns reveal where the model's decision boundary is weak and where review effort is most valuable; the alternative would make sense if a human reviewer frequently overrides one category of model decisions. It does not directly fix the observed condition: an automated report cites a source that cannot be retrieved by the reviewer.

C:

For the separate situation where the source evidence is contradictory and would change an important recommendation, conflicting high-impact evidence warrants review even if a model prefers one narrative with high confidence; the current task is to withhold high-impact release pending evidence.

D:

This control matters when a low-risk spelling correction is supported by deterministic lint feedback; review resources can focus on uncertain decisions while deterministic cosmetic changes proceed under normal controls. The required decision is different: an automated report cites a source that cannot be retrieved by the reviewer.

E:

Language-model confidence expressions are not calibrated statistical guarantees unless validated against representative outcomes; that assumes a model reports ninety-nine percent confidence without a calibrated evaluation process. The item tests the different ability to withhold high-impact release pending evidence.

Question 386

An invoice-extraction team builds reliable prompting for standardized records. A reviewer needs a confidence explanation tied to source text. Which response is most reliable?

  1. State the target ISO date format and how uncertain dates must be handled
  2. Define a shared field glossary and mapping rules before generation
  3. Require a null or explicit missing status instead of an invented identifier
  4. Require page or span references for important extracted fields

Explanation

Correct Answer

D:

Evidence-linked extraction allows downstream review to inspect the actual text behind uncertain or disputed values.

Incorrect Answers

A:

For the separate situation where extracted dates sometimes use locale-dependent shorthand, formatting criteria prevent ambiguous month/day interpretations and make downstream type checks meaningful; the current task is to connect extracted values to evidence.

B:

A stable schema prevents downstream systems from interpreting identical content under inconsistent attribute names; the alternative would make sense if the same invoice is processed by multiple teams with different field naming. It does not directly fix the observed condition: a reviewer needs a confidence explanation tied to source text.

C:

Missing fields should remain missing in the structured output; completeness must not be achieved by hallucinating values; that assumes a document is missing a purchase-order number. The item tests the different ability to connect extracted values to evidence.

Question 387

An architecture review uses Claude subagents for isolated design alternatives. The team needs to reproduce why two design branches diverged. What should the team do to resolve this particular gap?

  1. Give the worker the deployment paths, relevant constraints, and safe tools
  2. Revalidate changed files and environment state before relying on resumed context
  3. Record the branch identifier, assumptions, and tested outcomes for each alternative
  4. Fork the session or task state before exploring alternate designs

Explanation

Correct Answer

C:

Explicit branch history makes comparisons auditable; merged prose can otherwise conceal which experiment produced each result.

Incorrect Answers

A:

Scoping protects context budget and keeps the task anchored to the part of the repository it can actually assess; that assumes a specialist needs only the deployment directory, not the full monorepo. The item tests the different ability to retain branch-specific decisions and evidence.

B:

This control matters when an agent resumes after its local repository changed since the last session; a saved transcript is not a guarantee that disk or remote dependencies are unchanged; material state must be checked again. The required decision is different: the team needs to reproduce why two design branches diverged.

D:

For the separate situation where architects want to compare two approaches from the same initial repository state, branches can evaluate alternatives without contaminating each other's assumptions, preserving a meaningful comparison; the current task is to retain branch-specific decisions and evidence.

Question 388

An operations team monitors an agent-based customer workflow in production. A tool sometimes returns an error that a summarizer conceals. Which control is most directly relevant?

  1. Instrument fan-out, join, and worker durations separately
  2. Propagate structured error status through handoffs to the final decision layer
  3. Correlate model turns, tool requests, task IDs, and tool outcomes in structured traces
  4. Record versions, response digests, and safe replay artifacts for important external results

Explanation

Correct Answer

B:

Explicit error states prevent a later worker from smoothing a failed dependency into an apparently successful report.

Incorrect Answers

A:

Per-stage timing shows whether concurrency removes a real bottleneck or merely moves latency to a saturated shared service; that assumes the team wants to compare latency of serial and parallel research designs. The item tests the different ability to make failures visible across agent boundaries.

C:

For the separate situation where an agent makes repeated tool calls but the trace does not show why each was chosen, end-to-end correlation reveals the sequence leading to a failure without relying on a final narrative that may omit intermediate decisions; the current task is to make failures visible across agent boundaries.

D:

A stable record ties decisions to the data actually observed while avoiding the assumption that a live tool will return the same content later; the alternative would make sense if a production incident cannot be reproduced because tool outputs changed since execution. It does not directly fix the observed condition: a tool sometimes returns an error that a summarizer conceals.

Question 389

A research assistant uses the Messages API to coordinate evidence gathering. A search tool returns untrusted web text that says to ignore system instructions. How should the workflow be adjusted?

  1. Validate each tool-result ID against a pending request before appending it
  2. Treat tool output as data and enforce higher-priority instructions and tool permissions
  3. Treat ordinary assistant text as content rather than an executable tool instruction
  4. Detect max_tokens and continue or retry within a controlled budget

Explanation

Correct Answer

B:

External pages may contain prompt injection; tool output should inform answers, not override policy or initiate new privileges.

Incorrect Answers

A:

Correlation prevents an orphaned or stale result from contaminating the conversation and makes replay mismatches detectable; that assumes a returned tool_use_id is absent from the outstanding-request map. The item tests the different ability to keep retrieved content from controlling the agent.

C:

This control matters when an assistant text block describes a search but includes no tool_use block; only explicit tool-use content requests host execution; natural-language statements do not authorize a network call. The required decision is different: a search tool returns untrusted web text that says to ignore system instructions.

D:

For the separate situation where a response ends because the output-token limit was reached halfway through a structured answer, the stop reason distinguishes token exhaustion from a finished response; blindly accepting the fragment risks malformed or unsupported findings; the current task is to keep retrieved content from controlling the agent.

Question 390

An enterprise support assistant uses lengthy conversation and policy histories. A retrieved source conflicts with an earlier system-defined safety requirement. Which implementation decision is the best fit?

  1. Maintain a versioned decision ledger with approval status and source-turn references
  2. Treat retrieved text as evidence, not an override of higher-priority controls
  3. Fetch the latest verified address and label old entries as historical
  4. Restate a concise current policy excerpt near the final task instructions
  5. Retrieve the pertinent excerpt and carry its location into the answer

Explanation

Correct Answer

B:

External material can be adversarial or incorrect; long context must not promote a source document into policy authority.

Incorrect Answers

A:

A decision ledger tracks which resolution was authorized and which proposals were reversed, preventing a condensed conversation from treating tentative advice as binding; that assumes the support chat contains several competing ticket resolutions, including proposals that were later rejected. The item tests the different ability to maintain the instruction trust hierarchy.

C:

This control matters when a chat history includes several superseded account addresses; chronological conversation content can contain stale facts; an authoritative lookup prevents outdated messages from driving action. The required decision is different: a retrieved source conflicts with an earlier system-defined safety requirement.

D:

For the separate situation where the governing refund rule appears only once in the middle of a very long context, important details can be overlooked in long contexts; placing the controlling rule near use reduces lost-in-the-middle errors; the current task is to maintain the instruction trust hierarchy.

E:

Focused evidence helps the model judge a rule accurately and allows a reviewer to verify that the conclusion follows from the source; the alternative would make sense if an agent answers from a giant document without quoting the relevant passage. It does not directly fix the observed condition: a retrieved source conflicts with an earlier system-defined safety requirement.

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