Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 70

Question 690

When designing a customer refund workflow, the requirement is not satisfied unless out-of-policy exceptions receive an authorized decision and tool output is normalized before it enters the agent context. Which implementation addresses the two independent conditions?

  1. route the exception to a human approval step before execution; and use a PostToolUse processing hook to standardize the returned data
  2. implement only route the exception to a human approval step before execution; make no corresponding change for the second requirement
  3. let the agent approve an exception whenever its wording sounds confident; and preprocess only the original user message
  4. let the agent approve an exception whenever its wording sounds confident; and use a PostToolUse processing hook to standardize the returned data
  5. route the exception to a human approval step before execution; and preprocess only the original user message

Explanation

Correct Answer

A:

In a customer refund workflow, approval is a workflow boundary separate from language-model confidence; alongside this, post-execution interception can clean result fields and attach useful status without changing the model's tool selection, satisfying the distinct requirements that out-of-policy exceptions receive an authorized decision and that tool output is normalized before it enters the agent context.

Incorrect Answers

B:

In a customer refund workflow, approval is a workflow boundary separate from language-model confidence, but the independent mechanism for the other criterion is absent: post-execution interception can clean result fields and attach useful status without changing the model's tool selection; as written, the design does not ensure that tool output is normalized before it enters the agent context.

C:

In a customer refund workflow, model confidence is not delegated business authority and input normalization does not correct inconsistent tool outputs; the design would fail both conditions: out-of-policy exceptions receive an authorized decision and tool output is normalized before it enters the agent context.

D:

In a customer refund workflow, post-execution interception can clean result fields and attach useful status without changing the model's tool selection, but model confidence is not delegated business authority; therefore the separate requirement that out-of-policy exceptions receive an authorized decision would remain unsatisfied.

E:

In a customer refund workflow, approval is a workflow boundary separate from language-model confidence, but input normalization does not correct inconsistent tool outputs; therefore the separate requirement that tool output is normalized before it enters the agent context would remain unsatisfied.

Question 691

An audit of a source code audit helper asks how to make sure that a developer has private preferences that should not change team guidance. Which implementation should be selected?

  1. use CLAUDE.md imports to organize referenced instructions
  2. add private preferences to the shared repository CLAUDE.md
  3. audit overlapping CLAUDE.md files and resolve instruction contradictions
  4. place personal instructions in the user-level CLAUDE.md

Explanation

Correct Answer

D:

For a source code audit helper, global personal guidance is available to that user's sessions without committing preferences to the project; this directly establishes that a developer has private preferences that should not change team guidance instead of relying on an unrelated workflow control.

Incorrect Answers

A:

Although modular imports keep a short entry point while preserving reviewed details in separate files, that only supports the distinct goal that long project guidance is divided into maintainable source files, not the requirement that a developer has private preferences that should not change team guidance in a source code audit helper.

B:

In a source code audit helper, project instructions would impose one person's settings on the whole team; as a result, the stated requirement that a developer has private preferences that should not change team guidance would still not be met.

C:

Although clear ownership and scope prevent two directions from competing in the same coding task, that only supports the distinct goal that conflicting guidance from different files is detected before rollout, not the requirement that a developer has private preferences that should not change team guidance in a source code audit helper.

Question 692

The release gate for a large-codebase architecture explorer requires that reviewer feedback points to a specific missed requirement; it also requires that a tool receives a typed and bounded request. Which TWO actions meet the specified checks? Choose TWO.

  1. repeat the original prompt verbatim and expect a different outcome
  2. turn the feedback into a concrete test or change request
  3. accept one unstructured free-text argument for every operation
  4. define required properties and types in the input schema and validate arguments at the execution boundary

Explanation

Correct Answers

B:

For a large-codebase architecture explorer, actionable feedback narrows the next edit and makes improvement measurable, which directly establishes that reviewer feedback points to a specific missed requirement; the separate requirement that a tool receives a typed and bounded request is handled by a different action.

D:

For a large-codebase architecture explorer, input schemas describe legal arguments, while explicit validation prevents unsupported values from reaching tool logic, which directly establishes that a tool receives a typed and bounded request; the separate requirement that reviewer feedback points to a specific missed requirement is handled by a different action.

Incorrect Answers

A:

For a large-codebase architecture explorer, unchanged inputs do not target the identified deficiency, so this option cannot establish that reviewer feedback points to a specific missed requirement; correctly addressing a tool receives a typed and bounded request elsewhere does not repair that deficiency.

C:

For a large-codebase architecture explorer, unconstrained prose leaves parsing and validation errors to ad hoc handler inference, so this option cannot establish that a tool receives a typed and bounded request; correctly addressing reviewer feedback points to a specific missed requirement elsewhere does not repair that deficiency.

Question 693

For a cross-platform application migration, a reviewer identifies two independent obligations: a high-risk workflow only invokes specifically approved capabilities and two nested packages follow distinct framework conventions. Which proposal should be approved?

  1. implement only declare allowed-tools for the skill and enforce executor permissions; make no corresponding change for the second requirement
  2. declare allowed-tools for the skill and enforce executor permissions; and write one ambiguous instruction for both packages
  3. declare allowed-tools for the skill and enforce executor permissions; and create separate path-specific rule files for each package
  4. rely on a sentence asking the agent not to use risky tools; and create separate path-specific rule files for each package
  5. rely on a sentence asking the agent not to use risky tools; and write one ambiguous instruction for both packages

Explanation

Correct Answer

C:

In a cross-platform application migration, tool restrictions narrow the agent's operational capability while the skill runs; alongside this, scoped conventions avoid contradictory patterns between codebases sharing a monorepo, satisfying the distinct requirements that a high-risk workflow only invokes specifically approved capabilities and that two nested packages follow distinct framework conventions.

Incorrect Answers

A:

In a cross-platform application migration, tool restrictions narrow the agent's operational capability while the skill runs, but the independent mechanism for the other criterion is absent: scoped conventions avoid contradictory patterns between codebases sharing a monorepo; as written, the design does not ensure that two nested packages follow distinct framework conventions.

B:

In a cross-platform application migration, tool restrictions narrow the agent's operational capability while the skill runs, but the model must guess which framework convention applies; therefore the separate requirement that two nested packages follow distinct framework conventions would remain unsatisfied.

D:

In a cross-platform application migration, scoped conventions avoid contradictory patterns between codebases sharing a monorepo, but instruction-only boundaries cannot substitute for actual tool authorization; therefore the separate requirement that a high-risk workflow only invokes specifically approved capabilities would remain unsatisfied.

E:

In a cross-platform application migration, instruction-only boundaries cannot substitute for actual tool authorization and the model must guess which framework convention applies; the design would fail both conditions: a high-risk workflow only invokes specifically approved capabilities and two nested packages follow distinct framework conventions.

Question 694

In a technical review of a customer refund workflow, the architect must address the following constraints together: schema-valid output is also checked against source evidence; a conclusion supported by model agreement is also tested against source evidence. Which change set is most defensible?

  1. validate semantic constraints and provenance in application logic; and promote agreement alone to a guaranteed correctness signal
  2. assume a schema-conforming response is factually correct; and require source and business-rule checks regardless of model agreement
  3. assume a schema-conforming response is factually correct; and promote agreement alone to a guaranteed correctness signal
  4. validate semantic constraints and provenance in application logic; and require source and business-rule checks regardless of model agreement

Explanation

Correct Answer

D:

In a customer refund workflow, JSON shape does not prove that an extracted claim is supported by the document; alongside this, correlated model errors can survive consensus without external verification, satisfying the distinct requirements that schema-valid output is also checked against source evidence and that a conclusion supported by model agreement is also tested against source evidence.

Incorrect Answers

A:

In a customer refund workflow, JSON shape does not prove that an extracted claim is supported by the document, but shared biases can produce identical wrong conclusions; therefore the separate requirement that a conclusion supported by model agreement is also tested against source evidence would remain unsatisfied.

B:

In a customer refund workflow, correlated model errors can survive consensus without external verification, but syntactic conformance cannot guarantee source-level truth; therefore the separate requirement that schema-valid output is also checked against source evidence would remain unsatisfied.

C:

In a customer refund workflow, syntactic conformance cannot guarantee source-level truth and shared biases can produce identical wrong conclusions; the design would fail both conditions: schema-valid output is also checked against source evidence and a conclusion supported by model agreement is also tested against source evidence.

Question 695

Production readiness for a backend API maintenance team depends on ensuring that a small approved typo fix is completed without excessive planning and that independent stages of a complex job run efficiently. Select the most appropriate combined approach.

  1. launch a multi-stage planning effort for a single known line; and split the job into an arbitrary fixed chain of agents
  2. use direct execution for the limited reviewed edit; and divide work along independent outputs and run those stages in parallel
  3. launch a multi-stage planning effort for a single known line; and divide work along independent outputs and run those stages in parallel
  4. use direct execution for the limited reviewed edit; and split the job into an arbitrary fixed chain of agents

Explanation

Correct Answer

B:

In a backend API maintenance team, a bounded low-risk change does not require a long exploratory workflow; alongside this, dependency-aware decomposition avoids serial work when branches do not consume one another, satisfying the distinct requirements that a small approved typo fix is completed without excessive planning and that independent stages of a complex job run efficiently.

Incorrect Answers

A:

In a backend API maintenance team, extra ceremony adds latency without materially lowering risk and a rigid sequence creates waits even where no dependency exists; the design would fail both conditions: a small approved typo fix is completed without excessive planning and independent stages of a complex job run efficiently.

C:

In a backend API maintenance team, dependency-aware decomposition avoids serial work when branches do not consume one another, but extra ceremony adds latency without materially lowering risk; therefore the separate requirement that a small approved typo fix is completed without excessive planning would remain unsatisfied.

D:

In a backend API maintenance team, a bounded low-risk change does not require a long exploratory workflow, but a rigid sequence creates waits even where no dependency exists; therefore the separate requirement that independent stages of a complex job run efficiently would remain unsatisfied.

Question 696

In a technical review of a merchant support assistant, the architect must address the following constraints together: critical account identifiers survive a long support conversation; several requested client tools are returned in one response. Which change set is most defensible?

  1. rely only on a conversational summary of the earliest turns; and execute only the first returned tool call
  2. store verified IDs and decisions in an explicit structured case-facts record; and execute every tool_use block and return a result with each corresponding ID
  3. store verified IDs and decisions in an explicit structured case-facts record; and execute only the first returned tool call
  4. rely only on a conversational summary of the earliest turns; and execute every tool_use block and return a result with each corresponding ID
  5. implement only store verified IDs and decisions in an explicit structured case-facts record; make no corresponding change for the second requirement

Explanation

Correct Answer

B:

In a merchant support assistant, durable structured state prevents essential exact values being lost during summarization; alongside this, a multi-tool assistant turn can request independent actions and must receive all matching outcomes, satisfying the distinct requirements that critical account identifiers survive a long support conversation and that several requested client tools are returned in one response.

Incorrect Answers

A:

In a merchant support assistant, compression can omit or alter identifiers that later actions require and the unanswered tool uses remain unresolved and later reasoning loses evidence; the design would fail both conditions: critical account identifiers survive a long support conversation and several requested client tools are returned in one response.

C:

In a merchant support assistant, durable structured state prevents essential exact values being lost during summarization, but the unanswered tool uses remain unresolved and later reasoning loses evidence; therefore the separate requirement that several requested client tools are returned in one response would remain unsatisfied.

D:

In a merchant support assistant, a multi-tool assistant turn can request independent actions and must receive all matching outcomes, but compression can omit or alter identifiers that later actions require; therefore the separate requirement that critical account identifiers survive a long support conversation would remain unsatisfied.

E:

In a merchant support assistant, durable structured state prevents essential exact values being lost during summarization, but the independent mechanism for the other criterion is absent: a multi-tool assistant turn can request independent actions and must receive all matching outcomes; as written, the design does not ensure that several requested client tools are returned in one response.

Question 697

Tests for a customer refund workflow will pass only if an extractor always returns the identifiers required by downstream code and if reviewer changes remain auditable. Which proposed design directly satisfies both checks?

  1. declare mandatory schema fields in the required list; and overwrite an automated answer without preserving the reviewer decision
  2. treat every identifier as an optional free-form note; and overwrite an automated answer without preserving the reviewer decision
  3. declare mandatory schema fields in the required list; and log the reviewer identity, decision, justification, and artifact version
  4. treat every identifier as an optional free-form note; and log the reviewer identity, decision, justification, and artifact version

Explanation

Correct Answer

C:

In a customer refund workflow, required properties prevent accidental omission of key attributes; alongside this, a versioned approval trail supports diagnosis and later governance reviews, satisfying the distinct requirements that an extractor always returns the identifiers required by downstream code and that reviewer changes remain auditable.

Incorrect Answers

A:

In a customer refund workflow, required properties prevent accidental omission of key attributes, but the final state loses the reason it changed; therefore the separate requirement that reviewer changes remain auditable would remain unsatisfied.

B:

In a customer refund workflow, the consumer cannot reliably rely on a field that may be omitted and the final state loses the reason it changed; the design would fail both conditions: an extractor always returns the identifiers required by downstream code and reviewer changes remain auditable.

D:

In a customer refund workflow, a versioned approval trail supports diagnosis and later governance reviews, but the consumer cannot reliably rely on a field that may be omitted; therefore the separate requirement that an extractor always returns the identifiers required by downstream code would remain unsatisfied.

Question 698

For an order-status triage agent, a reviewer identifies two independent obligations: a payment-changing action obtains explicit user authorization and temporary service faults can be retried selectively. Which proposal should be approved?

  1. confirm the intended action and scope before invoking the write; and retry every error with the same request indefinitely
  2. treat a request for information as permission to mutate the account; and retry every error with the same request indefinitely
  3. confirm the intended action and scope before invoking the write; and include a retryable classification and bounded retry advice
  4. treat a request for information as permission to mutate the account; and include a retryable classification and bounded retry advice

Explanation

Correct Answer

C:

In an order-status triage agent, consent is specific to the side effect and cannot be inferred from a general inquiry; alongside this, typed transient failures allow backoff without treating permanent errors the same way, satisfying the distinct requirements that a payment-changing action obtains explicit user authorization and that temporary service faults can be retried selectively.

Incorrect Answers

A:

In an order-status triage agent, consent is specific to the side effect and cannot be inferred from a general inquiry, but permission and validation failures will not heal through repetition; therefore the separate requirement that temporary service faults can be retried selectively would remain unsatisfied.

B:

In an order-status triage agent, asking a question does not approve a financial change and permission and validation failures will not heal through repetition; the design would fail both conditions: a payment-changing action obtains explicit user authorization and temporary service faults can be retried selectively.

D:

In an order-status triage agent, typed transient failures allow backoff without treating permanent errors the same way, but asking a question does not approve a financial change; therefore the separate requirement that a payment-changing action obtains explicit user authorization would remain unsatisfied.

Question 699

During review of an engineering knowledge assistant, an unmet criterion states that a partially successful multi-agent job reports its limitations. Which measure resolves this specific gap?

  1. mark the whole request successful when any one branch returns data
  2. aggregate per-agent success and incomplete-state metadata
  3. return a typed failure status with its origin to the coordinator
  4. apply a bounded backoff only to classified retryable errors

Explanation

Correct Answer

B:

For an engineering knowledge assistant, partial status lets the coordinator qualify conclusions and retry missing branches; this directly establishes that a partially successful multi-agent job reports its limitations instead of relying on an unrelated workflow control.

Incorrect Answers

A:

In an engineering knowledge assistant, some required evidence may still be missing; as a result, the stated requirement that a partially successful multi-agent job reports its limitations would still not be met.

C:

Although an operational error must remain distinguishable from a valid empty result, that only supports the distinct goal that a failed retrieval agent does not masquerade as having found no matches, not the requirement that a partially successful multi-agent job reports its limitations in an engineering knowledge assistant.

D:

Although transient failures can heal without re-running permanently invalid requests, that only supports the distinct goal that a transient upstream outage is retried safely, not the requirement that a partially successful multi-agent job reports its limitations in an engineering knowledge assistant.

Question 700

In a developer productivity assistant, the design must ensure that dependent analysis begins only when source retrieval is complete and that a truncated response is not mistaken for successful completion. Choose TWO independent changes that directly meet these two requirements.

  1. handle max_tokens as truncation rather than end_turn
  2. run retrieval before analysis and pass its validated artifacts
  3. start the analysis and retrieval agents together without shared results
  4. interpret every non-tool stop reason as a successful final answer

Explanation

Correct Answers

A:

For a developer productivity assistant, reaching an output limit is a stop condition but does not establish task completion, which directly establishes that a truncated response is not mistaken for successful completion; the separate requirement that dependent analysis begins only when source retrieval is complete is handled by a different action.

B:

For a developer productivity assistant, a downstream agent needs actual upstream findings rather than an assumed future result, which directly establishes that dependent analysis begins only when source retrieval is complete; the separate requirement that a truncated response is not mistaken for successful completion is handled by a different action.

Incorrect Answers

C:

For a developer productivity assistant, the analysis branch cannot evaluate evidence it has not received, so this option cannot establish that dependent analysis begins only when source retrieval is complete; correctly addressing a truncated response is not mistaken for successful completion elsewhere does not repair that deficiency.

D:

For a developer productivity assistant, a length-limited result may be incomplete or unusable, so this option cannot establish that a truncated response is not mistaken for successful completion; correctly addressing dependent analysis begins only when source retrieval is complete elsewhere does not repair that deficiency.

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