Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 23

Question 220

An inference was repeated after a timeout but its earlier suggested action may already have executed. Which change would satisfy the constraint?

  1. Retry only failures considered transient
  2. Distinguish model-call retry from tool-action replay
  3. Use capped exponential backoff with jitter
  4. Centralize retry policy in shared middleware

Explanation

Correct Answer

B:

Separating generation retries from durable side-effect records avoids duplicate tool execution; the implementation must still recognize that model text has no transactional guarantee that would make action replays harmless.

Incorrect Answers

A:

A retry policy classifies failure causes and retries only the recoverable subset; this addresses a separate concern, not the fact that separating generation retries from durable side-effect records avoids duplicate tool execution; repeating all failures indiscriminately can amplify outages and costs.

C:

Randomized increasing delays spread load and respect retry guidance where supplied; the missing capability is more specific: separating generation retries from durable side-effect records avoids duplicate tool execution; unbounded retries create unpredictable latency and can hide an enduring fault.

D:

Shared policy consistently classifies errors and applies bounded retries across services; by comparison, the intended solution relies on this: separating generation retries from durable side-effect records avoids duplicate tool execution; it must still allow endpoint-specific rules for unsafe side effects.

Question 221

A summarizer needs to analyze a PDF rather than a screenshot of one paragraph. Which response is most defensible under these conditions?

  1. Pass an image as a typed image content block
  2. Use a supported file or document input method
  3. Choose a supported model identifier
  4. Validate MIME type and size before upload

Explanation

Correct Answer

B:

The integration supplies supported document content using the relevant API capability and access rules; the important constraint is that an arbitrary URL or filename does not automatically become readable source material.

Incorrect Answers

A:

Image blocks identify the media source while text instructions explain the intended analysis, whereas the requirement calls for another behavior: the integration supplies supported document content using the relevant API capability and access rules; putting a local path in ordinary text does not upload the image bytes to the API.

C:

The model field identifies the deployed model; production should verify current IDs and compatibility; the requirement instead turns on this behavior: the integration supplies supported document content using the relevant API capability and access rules; guessing an informal nickname can yield an invalid-model error or unexpected alias changes.

D:

Server-side checks bound resource use and ensure supported input formats before model submission; this cannot substitute for the behavior that the integration supplies supported document content using the relevant API capability and access rules, because client-provided extensions and names are not reliable content validation alone.

Question 222

A regulated payment workflow accepts generated amount changes, requires an auditable explanation of the recommendation, and must leave no partial ledger update if a later write fails. Which THREE server controls address the validation, traceability, and atomicity requirements? Choose THREE.

  1. Validate generated action arguments server-side
  2. Treat a retrieved webpage as untrusted data
  3. Record immutable provenance for sensitive decisions
  4. Use transaction controls for critical writes
  5. Require explicit confirmation for irreversible actions
  6. Audit privileged changes with actor attribution

Explanation

Correct Answers

A:

Backend checks enforce type, range, scope, and approval before any mutation; notably, the model's fluent output is not an authoritative business validation result.

C:

Audit metadata preserves input references, applicable policy, actor, and versioned output, subject to this requirement: a final answer without source history is not adequate for reconstructing the decision.

D:

Atomic transactions or compensating workflows preserve business consistency across approved changes; the implementation must still recognize that model reasoning does not guarantee database rollback.

Incorrect Answers

B:

External content is evidence, not an instruction source allowed to override the application's policies; in contrast, backend checks enforce type, range, scope, and approval before any mutation, while prompt labeling helps but tool-side access restrictions must remain enforced.

E:

A confirmation gate makes the consequential action a separate authorized decision; the missing capability is more specific: backend checks enforce type, range, scope, and approval before any mutation; interpreting ordinary conversational intent as final approval can cause unintended deletion.

F:

An audit trail connects the principal, approved operation, target, and result; by comparison, the intended solution relies on this: backend checks enforce type, range, scope, and approval before any mutation; a text summary by the model is not sufficient evidence of a committed security change.

Question 223

Many clients receive throttling at once and an immediate retry would synchronize their next attempt. Which choice should be used here?

  1. Attach an idempotency token to a business write
  2. Distinguish model-call retry from tool-action replay
  3. Use capped exponential backoff with jitter
  4. Evaluate semantic behavior with labeled fixtures

Explanation

Correct Answer

C:

Randomized increasing delays spread load and respect retry guidance where supplied; in production, unbounded retries create unpredictable latency and can hide an enduring fault.

Incorrect Answers

A:

Stable operation identity lets the transactional endpoint suppress duplicate side effects; here, the controlling capability is that randomized increasing delays spread load and respect retry guidance where supplied, but randomizing the token on every retry defeats the protection.

B:

Separating generation retries from durable side-effect records avoids duplicate tool execution; the missing capability is more specific: randomized increasing delays spread load and respect retry guidance where supplied; model text has no transactional guarantee that would make action replays harmless.

D:

A test set with expected outcomes measures content quality, not only code execution; the needed capability is different—randomized increasing delays spread load and respect retry guidance where supplied—and a green HTTP response cannot establish answer correctness.

Question 224

Many apps call Claude and all must apply the same authentication, budgets, and logging policies. Which solution should the engineer adopt?

  1. Use a wrapper service for shared organizational controls
  2. Use a typed abstraction for business-specific inference
  3. Use server-sent events for streamed responses
  4. Use a raw HTTPS REST client when necessary

Explanation

Correct Answer

A:

A centralized proxy can enforce cross-cutting organizational controls before forwarding requests—with this limitation: the proxy also becomes a reliability and credential-protection responsibility.

Incorrect Answers

B:

A domain wrapper hides model-request details behind validated inputs and outputs; this cannot substitute for the behavior that a centralized proxy can enforce cross-cutting organizational controls before forwarding requests, because abstractions should expose meaningful errors instead of swallowing model uncertainty.

C:

SSE carries incremental events from the service to the client as generation proceeds; in contrast, a centralized proxy can enforce cross-cutting organizational controls before forwarding requests, while it is not a general-purpose bidirectional client messaging transport.

D:

Direct REST requests can call the same documented endpoint with correct headers and body; this addresses a separate concern, not the fact that a centralized proxy can enforce cross-cutting organizational controls before forwarding requests; the developer must implement encoding, retries, and response parsing independently.

Question 225

A retry handler must be tested under specific stop reasons without paying for live API calls. Which of these alternatives serves the stated goal?

  1. Evaluate semantic behavior with labeled fixtures
  2. Run contract tests against a versioned API schema
  3. Use a safe degraded response on dependency outage
  4. Mock model responses for deterministic unit tests

Explanation

Correct Answer

D:

A mock response supplies controlled payloads so application logic can be tested repeatably; a crucial qualification is that it does not prove a real model will produce the same semantic output.

Incorrect Answers

A:

A test set with expected outcomes measures content quality, not only code execution; in contrast, a mock response supplies controlled payloads so application logic can be tested repeatably, while a green HTTP response cannot establish answer correctness.

B:

Contract tests enforce agreed field names, types, and error shapes across producer and client; this addresses a separate concern, not the fact that a mock response supplies controlled payloads so application logic can be tested repeatably; casual manual testing may miss a breaking field change.

C:

A fallback can queue the request or return a limited verified response without fabricating model results; here, the controlling capability is that a mock response supplies controlled payloads so application logic can be tested repeatably, but retrying indefinitely inside the user's HTTP connection harms availability.

Question 226

The model requests an operation and the dispatcher must identify the invocation it is about to fulfill. Which action most directly closes the gap?

  1. Return a matching tool_result block
  2. Reconcile all returned tool results before synthesis
  3. Report tool failures explicitly
  4. Read the returned tool_use ID and name

Explanation

Correct Answer

D:

Each client tool request includes metadata that ties subsequent execution and results to that call; guessing an operation from nearby text can dispatch the wrong function.

Incorrect Answers

A:

The following turn includes a result linked by tool_use_id so the model can continue coherently; here, the controlling capability is that each client tool request includes metadata that ties subsequent execution and results to that call, but returning an unrelated human message loses the execution association.

B:

The harness maps all available results to their calls before requesting a comparative answer; the missing capability is more specific: each client tool request includes metadata that ties subsequent execution and results to that call; ignoring a pending result can produce an unsupported recommendation.

C:

A structured error result communicates the failed operation and supports safe recovery; by comparison, the intended solution relies on this: each client tool request includes metadata that ties subsequent execution and results to that call; silently replacing an error with an empty success encourages false completion claims.

Question 227

A malicious customer adds a persuasive instruction to transfer another user's order data. What should be implemented to solve this problem?

  1. Treat a retrieved webpage as untrusted data
  2. Authorize each requested operation at the resource level
  3. Verify tool permissions independently of text
  4. Filter outgoing sensitive data at the egress layer

Explanation

Correct Answer

C:

Server-side authorization checks bind each requested resource to the authenticated principal; the important constraint is that a system-prompt ban on disclosure cannot replace real access control.

Incorrect Answers

A:

External content is evidence, not an instruction source allowed to override the application's policies; the needed capability is different—server-side authorization checks bind each requested resource to the authenticated principal—and prompt labeling helps but tool-side access restrictions must remain enforced.

B:

Access policies check whether that principal may perform the exact requested change on the target object, whereas the requirement calls for another behavior: server-side authorization checks bind each requested resource to the authenticated principal; successful login alone does not grant permission across all records.

D:

Deterministic output and action checks block unauthorized exfiltration before it leaves the trusted environment; the requirement instead turns on this behavior: server-side authorization checks bind each requested resource to the authenticated principal; model self-policing alone cannot guarantee detection of every malicious instruction.

Question 228

A new verified support ticket should launch classification immediately when a separate system emits a creation event. Which option best fits the described workflow?

  1. Use an API gateway for request controls
  2. Use an event-driven trigger for external updates
  3. Use an explicit tool definition for callable business actions
  4. Use a dedicated tool adapter for private systems

Explanation

Correct Answer

B:

Event handlers initiate the approved model workflow on a real business change; the important constraint is that polling or repeated user prompts are unnecessary when a reliable event is available.

Incorrect Answers

A:

A gateway applies uniform boundary policies before passing requests to workers, whereas the requirement calls for another behavior: event handlers initiate the approved model workflow on a real business change; a gateway alone does not verify every downstream tool's resource-level authorization.

C:

Tool schemas advertise an available operation and define its structured inputs; the requirement instead turns on this behavior: event handlers initiate the approved model workflow on a real business change; a prose instruction mentioning a function does not register it with the API.

D:

An adapter exposes narrowly typed operations and translates requests into the legacy interface; this cannot substitute for the behavior that event handlers initiate the approved model workflow on a real business change, because the model does not acquire direct private-network access by seeing the system's name.

Question 229

The HTTP request succeeds but a customer categorization is factually wrong or lacks a required field. Which mechanism should the team prioritize?

  1. Trace the external tool service
  2. Quarantine malformed output for review
  3. Inspect the serialized outgoing payload
  4. Inspect parsed model content separately

Explanation

Correct Answer

D:

Output-level checks isolate semantic and schema failures from network transport errors; in production, treating every wrong answer as an API outage conceals prompt and evaluation problems.

Incorrect Answers

A:

A distributed trace follows the tool dispatcher to the failing downstream dependency; this addresses a separate concern, not the fact that output-level checks isolate semantic and schema failures from network transport errors; rewriting system instructions will not repair an unavailable invoice endpoint.

B:

Failed artifacts are retained securely for diagnostics and excluded from the production write path; here, the controlling capability is that output-level checks isolate semantic and schema failures from network transport errors, but automatically coercing unknown fields risks corrupting downstream ledgers.

C:

Logging a redacted request reveals malformed fields introduced by serialization or adapters; the missing capability is more specific: output-level checks isolate semantic and schema failures from network transport errors; reading the model's final natural-language answer cannot reveal a request that never reached generation.

Question 230

A procurement UI displays a proposed purchase order awaiting final approval. How should the implementation handle this need?

  1. Clearly distinguish a draft from a committed action
  2. Separate display text from operational commands
  3. Expose a safe retry action after transient faults
  4. Provide an escalation channel for uncertain outcomes

Explanation

Correct Answer

A:

Labels and separate commit controls prevent users from assuming a draft has been submitted; model-generated wording should not make a simulated operation appear executed.

Incorrect Answers

B:

The service distinguishes a human-readable recommendation from an authorized tool action, whereas the requirement calls for another behavior: labels and separate commit controls prevent users from assuming a draft has been submitted; parsing ordinary prose to infer permission for a write is unsafe.

C:

The UI can offer a bounded retry with clear status rather than silently losing the task; however, the critical mechanism here is that labels and separate commit controls prevent users from assuming a draft has been submitted, and the backend must still prevent duplicated side effects when retrying.

D:

Escalation transfers context to an authorized person instead of fabricating a final decision; the requirement instead turns on this behavior: labels and separate commit controls prevent users from assuming a draft has been submitted; human review must be a real operational path, not only a sentence in a prompt.

Claude Certified Developer - Foundations (CCDV-F) Practice Tests