Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 26

Question 250

A recursive support agent repeatedly asks for new tools without approaching a resolution. Which host control prevents unbounded execution?

  1. Advance the paused assistant turn with its response content preserved
  2. Send the refusal through the approved response and escalation policy
  3. Safeguard the tool implementation in the authorized execution environment
  4. Enforce a step budget and a defined fallback when the loop exceeds it

Explanation

Correct Answer

D:

Independent limits on turns, time, and side effects prevent runaway agent cycles. Exceeding a budget should trigger a controlled handoff or incomplete status, not infinite recursion.

Incorrect Answers

A:

A pause_turn can mark a server-tool iteration boundary; that serves continuing a paused server-tool loop, whereas the case requires bounding an autonomous agent loop.

B:

Because a refusal is a distinct termination state, the measure supports handling refusal as a distinct state rather than the needed bounding an autonomous agent loop.

C:

Client tools are interfaces advertised to the model; the hosting application runs their operations; its benefits for distinguishing model intent from tool execution do not resolve the need for bounding an autonomous agent loop.

Question 251

A flaky AI review blocks every PR even when the model's response is malformed. What control should be added?

  1. Validate the review payload and apply a documented failure policy
  2. Seek machine-readable JSON output and validate its schema
  3. Call for file path, line range and finding rationale in a validated result schema
  4. Constrain retrieval and review to the changed files and their affected dependencies
  5. Insist on verifiable evidence before elevating the claim to a blocking finding

Explanation

Correct Answer

A:

CI must distinguish tooling errors from confirmed code defects. Schema checks and fail-open or fail-closed rules selected by risk prevent random parsing failures from masquerading as findings.

Incorrect Answers

B:

Structured output is easier for CI to parse and gate deterministically; a method suited to parsing review output safely in CI does not meet the need for handling invalid AI-review output in a pipeline.

C:

For generating location-specific PR findings, location-aware structured findings allow a bot to attach actionable comments; however, this requirement concerns handling invalid AI-review output in a pipeline.

D:

A diff-aware prompt and bounded context reduce cost while preserving relevant evidence; that may serve constraining CI review to changed scope, but it is no substitute for handling invalid AI-review output in a pipeline.

E:

Model assertions are hypotheses until supported by inspected code or tests; this may support verifying CI review claims before blocking but cannot alone meet the requirement for handling invalid AI-review output in a pipeline.

Question 252

A feature request affects authentication across many modules and requirements are still unclear. Which mode should Claude Code use first?

  1. Provide the plan for approval before allowing write-capable execution
  2. Perform the narrow approved edit directly and run the relevant check
  3. Start in plan mode for investigation and architectural approval
  4. Include the architectural trade-off and request the missing business decision

Explanation

Correct Answer

C:

Plan mode helps explore the codebase and propose changes without immediately applying edits. It is suited to ambiguous, high-impact work where a premature patch would be risky.

Incorrect Answers

A:

Although sensitive migrations benefit from separating design review from mutation, the choice addresses requiring plan approval for a migration and not the requested planning before a cross-cutting implementation.

B:

A bounded, well-specified low-risk change often needs no lengthy planning phase; the deciding condition is planning before a cross-cutting implementation, whereas this action concerns implementing a tiny well-scoped change.

D:

An unresolved design choice is not safe to infer from code alone; this addresses resolving architecture ambiguity before editing, while the question is about planning before a cross-cutting implementation.

Question 253

A project team wants a reviewer that cannot alter the code being assessed. Which agent role separation is safest?

  1. Send each worker only the minimum task-specific context and authorized tools
  2. Assign analysis-only capabilities to the reviewer and editing tools to a separate implementer
  3. Hand off cross-specialty requests through an orchestrator with an explicit task contract
  4. Decompose independent expert analyses and reserve integration for the orchestrator
  5. Require completed evidence review before synthesis

Explanation

Correct Answer

B:

Tool partitioning enforces role boundaries. A reviewer that shares the author's write privileges can silently modify the evidence it is supposed to judge.

Incorrect Answers

A:

When considering compartmentalizing delegated context and privileges, specialization should come with least-necessary information and access; the requirement for separating creator and reviewer privileges remains unmet.

C:

Hub-and-spoke mediation controls ownership, context, and failure propagation; minimizing unpredictable peer-to-peer dependencies and separating creator and reviewer privileges require different controls in this situation.

D:

Focused workers can receive targeted instructions and smaller context windows; the unsatisfied requirement is separating creator and reviewer privileges, not matching specialist agents to distinct competencies.

E:

A dependency edge enforces that the reporting worker receives validated source material; in this scenario, enforcing a dependency before synthesis is secondary to separating creator and reviewer privileges.

Question 254

A security scan flags a vulnerable dependency but the agent edits source before checking installed versions. What is a better first action?

  1. Employ repository search to locate relevant symbols and then read focused files
  2. Inspect dependency manifests and lockfiles to confirm the affected version
  3. Locate references and consumer contracts before changing the shared definition
  4. Unify a short verified configuration summary with exact file references

Explanation

Correct Answer

B:

Current local package metadata grounds the remediation plan. Assuming a particular library version can lead to inappropriate API changes or false-positive vulnerability fixes.

Incorrect Answers

A:

Search-first exploration narrows the context to likely dependencies; the action supports narrowing large-repository inspection by search but falls short of grounding dependency remediation in repository metadata in this case.

C:

Shared APIs can affect distant packages; that addresses mapping consumers of a shared interface, whereas this case calls for grounding dependency remediation in repository metadata.

D:

Shared grounded notes prevent repeated expensive reads while retaining the ability to verify details; this concerns reusing concise verified repository findings, leaving the requirement for grounding dependency remediation in repository metadata outstanding.

Question 255

A custom command produces inconsistent output sections because its acceptance format is vague. What should its instructions include?

  1. Formalize a custom slash-invocable command or skill for the PR summary
  2. Delete embedded secrets and access approved credentials through secure runtime configuration
  3. Keep supporting resources within the skill and load them when needed
  4. Define required headings, evidence fields and a completion checklist
  5. Supply the skill a clear trigger description and intended use cases

Explanation

Correct Answer

D:

A bounded output contract makes a reusable workflow easier to verify. Without one, the same command can return materially different artifacts across runs.

Incorrect Answers

A:

Named command workflows provide repeatable behavior for a familiar task; its benefits for creating a reusable named PR-summary action do not resolve the need for standardizing a command's output shape.

B:

Skill contents may be loaded into model context and shared in repositories; supporting securing reusable skill contents cannot replace the required standardizing a command's output shape.

C:

Skills can bundle reference material and executable helpers for a focused task; this contributes to packaging on-demand skill resources without meeting the need for standardizing a command's output shape.

E:

Because descriptive metadata improves discovery and correct invocation, the measure supports improving skill discoverability rather than the needed standardizing a command's output shape.

Question 256

A prompt uses only examples with one correct entity and fails when multiple entities appear. What expansion improves robustness?

  1. Expose diverse input date formats paired with one required normalized representation
  2. Adopt varied fictional entities and emphasize that examples illustrate structure, not actual values
  3. Log example-set versions and compare outcomes on a fixed test suite
  4. Demonstrate zero, one and multiple valid extraction matches as appropriate

Explanation

Correct Answer

D:

Cardinality examples teach when to return arrays versus absent values. They prevent an implicit assumption that every document contains exactly one match.

Incorrect Answers

A:

For demonstrating cross-locale date normalization, contrasting examples demonstrate conversion rules while preserving a single target shape; however, this requirement concerns calibrating extraction cardinality.

B:

Reusing the same names across all demonstrations can invite copying; that may serve preventing sample values from leaking into outputs, but it is no substitute for calibrating extraction cardinality.

C:

Versioned few-shot experiments reveal the impact of changes; this may support measuring few-shot changes with controlled evaluations but cannot alone meet the requirement for calibrating extraction cardinality.

Question 257

A reporting workflow starts a large analyst agent for a task needing only a simple status lookup. What orchestration refinement makes sense?

  1. Remove duplicate entries from planned search tasks and share a normalized evidence index
  2. Constrain delegation depth and concurrent subagent count at the orchestration layer
  3. Route trivial deterministic steps directly and reserve specialists for nontrivial reasoning
  4. Perform independent read-only specialist tasks concurrently and join their outputs

Explanation

Correct Answer

C:

Agent delegation has token, latency and observability costs. A coordinator should use basic tools for routine lookups and invoke agents only when specialized judgment adds value.

Incorrect Answers

A:

Centralizing a work plan and results catalogue prevents workers from independently repeating identical queries; the option would serve avoiding redundant delegated searches, not avoiding needless specialist agent invocations as specified in the stem.

B:

Limits on nesting and worker breadth bound cost and latency; this addresses constraining recursive multi-agent delegation, while the question is about avoiding needless specialist agent invocations.

D:

Although when branches have no data dependency, parallel execution reduces waiting time while leaving the coordinator responsible for combining results and handling any branch failure, the choice addresses parallelizing independent evidence gathering and not the requested avoiding needless specialist agent invocations.

Question 258

A downstream report generator receives an error object where it expects a structured analysis. How should the handoff contract change?

  1. Preserve a partial workflow outcome and identify the unavailable branch explicitly
  2. Classify separately tool failure from a valid search with zero matches
  3. Present completed, failed and pending work as separate outcome categories
  4. Use tagged success and failure result types with explicit branching

Explanation

Correct Answer

D:

A discriminated result lets each consumer handle a failure without parsing prose as data. It also prevents a tool diagnostic from being summarized as a factual finding.

Incorrect Answers

A:

Partial success is not whole-task success; preserving partial-failure status across worker aggregation and making errors explicit in worker result schemas require different controls in this situation.

B:

Infrastructure errors must propagate as errors, not negative evidence; the unsatisfied requirement is making errors explicit in worker result schemas, not propagating search errors without treating them as no-match evidence.

C:

Progress counters must reflect task outcomes rather than merely finished attempts; in this scenario, distinguishing task completion from task success is secondary to making errors explicit in worker result schemas.

Question 259

A model returns syntactically valid JSON but a fabricated invoice total. Which statement about structured outputs is correct?

  1. Schema validity must be complemented by source-based semantic validation
  2. Specify a nullable field explicitly and use a documented missing-value convention
  3. Focus on simplifying the schema to supported constructs and validate remaining rules in application code
  4. Employ nested object definitions with typed address properties

Explanation

Correct Answer

A:

A schema constrains representation, not factual truth. Numerical fields can be correctly typed yet unsupported by the document or calculation.

Incorrect Answers

B:

A deliberate null representation lets consumers distinguish unknown data from malformed output; this concerns representing legitimate missing data in a schema, leaving the requirement for distinguishing schema conformance from factual accuracy outstanding.

C:

Constrained decoding has a documented supported JSON Schema subset; respecting structured-output schema limitations and distinguishing schema conformance from factual accuracy are distinct responsibilities under these constraints.

D:

A nested schema preserves structural relationships and makes each field independently verifiable; the action supports constraining nested JSON extraction structures but falls short of distinguishing schema conformance from factual accuracy in this case.

Question 260

A customer claims an urgent exception to a standard no-refund policy. Which safe action preserves service quality?

  1. Disambiguate the intended action and confirm authorization before choosing either operation
  2. Transfer to the security or fraud workflow and suspend sensitive actions
  3. Capture the exception rationale and route it through the authorized review process
  4. Seek authorized identity clarification before account-specific action

Explanation

Correct Answer

C:

Unusual circumstances may justify discretion but need a human or policy-sanctioned decision. Automatic denial or approval without the documented exception path may be inappropriate.

Incorrect Answers

A:

Destructive and reversible operations are not interchangeable; this contributes to clarifying an ambiguous destructive request without meeting the need for processing policy exceptions through approval.

B:

Fraud indicators change the permitted handling of customer accounts; that serves routing suspected fraud out of standard support, whereas the case requires processing policy exceptions through approval.

D:

Missing identity is a safety-critical ambiguity; its benefits for escalating unresolved customer identity do not resolve the need for processing policy exceptions through approval.

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