Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 3

Question 20

A token economics review requires two independent, explicitly defined outcomes: the first workstream must explain why a plausible answer can still contain an unsupported factual claim; a second must brainstorm diverse naming candidates without demanding identical wording on every run. Which TWO actions directly provide these capabilities? Choose TWO.

  1. Use one representative example
  2. Use contrasting few-shot examples
  3. Recognize next-token generation
  4. Use varied sampling only for ideation
  5. Use low-temperature generation for repeatable extraction

Explanation

Correct Answers

C:

The model predicts likely continuations conditioned on context rather than querying an authoritative truth database by default; this directly supports task 1 to explain why a plausible answer can still contain an unsupported factual claim, not the separate task or tasks to brainstorm diverse naming candidates without demanding identical wording on every run; fluent prose is not evidence that a source was verified.

D:

Greater sampling diversity can produce a wider range of suggestions for exploratory work; this directly supports task 2 to brainstorm diverse naming candidates without demanding identical wording on every run, not the separate task or tasks to explain why a plausible answer can still contain an unsupported factual claim; it is not appropriate as the sole correctness control for structured production data.

Incorrect Answers

A:

A single demonstration can anchor the expected shape and tone of a generated response; this implements the different outcome to clarify a specific formatting convention that a plain instruction was not conveying rather than the required tasks to explain why a plausible answer can still contain an unsupported factual claim and to brainstorm diverse naming candidates without demanding identical wording on every run, and an example is not a formal guarantee that every response will conform, so it is not a direct solution to any of the assigned workstreams.

B:

Multiple representative cases help define a decision boundary and reduce ambiguity; this implements the different outcome to distinguish borderline policy labels with examples of both allowed and disallowed cases rather than the required tasks to explain why a plausible answer can still contain an unsupported factual claim and to brainstorm diverse naming candidates without demanding identical wording on every run, and repeating near-identical samples does not cover edge cases, so it is not a direct solution to any of the assigned workstreams.

E:

Lower randomness often improves consistency for narrow, constrained tasks; this implements the different outcome to reduce variation when a service maps documents into consistent categories rather than the required tasks to explain why a plausible answer can still contain an unsupported factual claim and to brainstorm diverse naming candidates without demanding identical wording on every run, and it does not guarantee identical output or replace schema validation, so it is not a direct solution to any of the assigned workstreams.

Question 21

Two separate tickets in an incident response backlog require different controls. One needs to prevent a slow external inventory service from blocking an agent indefinitely; the other needs to recover from a failed search while preserving the agent's chance to choose another path. Which implementation pair is correct?

  1. First component: Set a tool execution timeout. Second component: Return a structured tool error to the model.
  2. First component: Keep user-specific state out of a shared agent process. Second component: Use the Claude Agent SDK for an integrated tool-driven agent.
  3. First component: Set a tool execution timeout. Second component: Keep user-specific state out of a shared agent process.
  4. First component: Use the Claude Agent SDK for an integrated tool-driven agent. Second component: Return a structured tool error to the model.

Explanation

Correct Answer

A:

The first proposal fits its assigned role because deadlines bound latency and convert stalled calls into explicit recoverable failures, while the second proposal fits its assigned role because tool-result errors surface the failure without impersonating a successful action; their respective acceptance criteria are to prevent a slow external inventory service from blocking an agent indefinitely and to recover from a failed search while preserving the agent's chance to choose another path.

Incorrect Answers

B:

The first proposal serves the different purpose to serve many tenants without mixing their tool outputs or prior messages instead of the assigned first role because per-request identity and session boundaries prevent cross-tenant context leakage, while the second proposal serves the different purpose to build a coding agent that needs managed tool execution, sessions, and event handling instead of the assigned second role because the SDK provides agent-loop primitives and integrations that remove much of the custom harness work; their respective acceptance criteria are to prevent a slow external inventory service from blocking an agent indefinitely and to recover from a failed search while preserving the agent's chance to choose another path.

C:

The first proposal fits its assigned role because deadlines bound latency and convert stalled calls into explicit recoverable failures, while the second proposal serves the different purpose to serve many tenants without mixing their tool outputs or prior messages instead of the assigned second role because per-request identity and session boundaries prevent cross-tenant context leakage; their respective acceptance criteria are to prevent a slow external inventory service from blocking an agent indefinitely and to recover from a failed search while preserving the agent's chance to choose another path.

D:

The first proposal serves the different purpose to build a coding agent that needs managed tool execution, sessions, and event handling instead of the assigned first role because the SDK provides agent-loop primitives and integrations that remove much of the custom harness work, while the second proposal fits its assigned role because tool-result errors surface the failure without impersonating a successful action; their respective acceptance criteria are to prevent a slow external inventory service from blocking an agent indefinitely and to recover from a failed search while preserving the agent's chance to choose another path.

Question 22

A design meeting for a customer platform platform concludes that two isolated services require different behavior: one must extract claims from a document while retaining its visual layout, and another must recover a response that ended because its output budget was exhausted. Which arrangement meets the specification?

  1. First component: Use supported document or PDF input. Second component: Handle message_stop and final usage events.
  2. First component: Include an image content block. Second component: Handle the max_tokens stop_reason.
  3. First component: Use supported document or PDF input. Second component: Handle the max_tokens stop_reason.
  4. First component: Handle message_stop and final usage events. Second component: Include an image content block.

Explanation

Correct Answer

C:

The first proposal fits its assigned role because document-aware input lets a capable model consider page content and graphics when supported, while the second proposal fits its assigned role because the finish reason identifies truncation, so a caller may need a continuation or a revised token budget; their respective acceptance criteria are to extract claims from a document while retaining its visual layout and to recover a response that ended because its output budget was exhausted.

Incorrect Answers

A:

The first proposal fits its assigned role because document-aware input lets a capable model consider page content and graphics when supported, while the second proposal serves the different purpose to record complete streamed output and token usage accurately instead of the assigned second role because the final stream events signal completion and deliver accumulated metadata beyond individual text fragments; their respective acceptance criteria are to extract claims from a document while retaining its visual layout and to recover a response that ended because its output budget was exhausted.

B:

The first proposal serves the different purpose to analyze a photographed shipping label with the supported multimodal Messages input instead of the assigned first role because image inputs are supplied as typed content with appropriate source data or references, while the second proposal fits its assigned role because the finish reason identifies truncation, so a caller may need a continuation or a revised token budget; their respective acceptance criteria are to extract claims from a document while retaining its visual layout and to recover a response that ended because its output budget was exhausted.

D:

The first proposal serves the different purpose to record complete streamed output and token usage accurately instead of the assigned first role because the final stream events signal completion and deliver accumulated metadata beyond individual text fragments, while the second proposal serves the different purpose to analyze a photographed shipping label with the supported multimodal Messages input instead of the assigned second role because image inputs are supplied as typed content with appropriate source data or references; their respective acceptance criteria are to extract claims from a document while retaining its visual layout and to recover a response that ended because its output budget was exhausted.

Question 23

A production API review requires two independent, explicitly defined outcomes: the first workstream must detect unsafe permissions and schema changes before merging an agent tool; a second must maintain multiple Claude entry points without drifting retry and logging behavior. Which TWO actions directly provide these capabilities? Choose TWO.

  1. Use asynchronous I/O for concurrent requests
  2. Use exponential backoff with jitter
  3. Use pull requests and focused code review
  4. Refactor duplicated request orchestration
  5. Use feature flags for risky behavior changes

Explanation

Correct Answers

C:

Peer review creates an explicit checkpoint for correctness, security, and maintainability; this directly supports task 1 to detect unsafe permissions and schema changes before merging an agent tool, not the separate task or tasks to maintain multiple Claude entry points without drifting retry and logging behavior; running a linter alone cannot assess authorization intent.

D:

A shared, tested client layer centralizes cross-cutting concerns and reduces inconsistent fixes; this directly supports task 2 to maintain multiple Claude entry points without drifting retry and logging behavior, not the separate task or tasks to detect unsafe permissions and schema changes before merging an agent tool; copying the same networking code into every route magnifies defects.

Incorrect Answers

A:

Async execution overlaps I/O-bound tasks while maintaining control over concurrency; this implements the different outcome to serve independent API calls without blocking a worker thread for each network wait rather than the required tasks to detect unsafe permissions and schema changes before merging an agent tool and to maintain multiple Claude entry points without drifting retry and logging behavior, and it does not by itself make CPU-bound model processing faster, so it is not a direct solution to any of the assigned workstreams.

B:

Increasing randomized delays reduce contention and respect service throttling signals; this implements the different outcome to recover from transient 429 responses without synchronized retry storms rather than the required tasks to detect unsafe permissions and schema changes before merging an agent tool and to maintain multiple Claude entry points without drifting retry and logging behavior, and immediate identical retries can extend an overload event, so it is not a direct solution to any of the assigned workstreams.

E:

Runtime controls can segment rollout and permit quick disabling without a code rebuild; this implements the different outcome to activate a new tool execution path for only internal testers rather than the required tasks to detect unsafe permissions and schema changes before merging an agent tool and to maintain multiple Claude entry points without drifting retry and logging behavior, and hard-coded unconditional activation exposes every tenant simultaneously, so it is not a direct solution to any of the assigned workstreams.

Question 24

During acceptance testing for an MCP interoperability service, one requirement is to let an assistant recover from a missing order without inventing its details. What should the developer implement?

  1. Map tool_use IDs to results
  2. Return structured tool errors
  3. Write task-specific tool descriptions
  4. Separate read tools from write tools

Explanation

Correct Answer

B:

Machine-readable failures distinguish not-found, permission denied, and transient errors for appropriate follow-up; this behavior directly supports the need to let an assistant recover from a missing order without inventing its details. Pretending failure is an empty successful record hides operational meaning.

Incorrect Answers

A:

This approach serves the separate need to support multiple simultaneous lookups without mixing their responses because correlation IDs connect each observation to the exact invocation that requested it; it does not establish the outcome to let an assistant recover from a missing order without inventing its details, and pairing results by arrival order is unsafe under concurrency.

C:

For the alternate objective to let Claude distinguish a read-only invoice lookup from an invoice update, clear capabilities and side effects guide tool selection without granting new authority; the feature here must let an assistant recover from a missing order without inventing its details, and vague overlapping descriptions increase erroneous calls.

D:

Distinct interfaces simplify least-privilege grants and side-effect controls; this applies to the different goal to require stricter approval for actions that change customer data, not the requirement to let an assistant recover from a missing order without inventing its details; a generic execute-anything function makes authorization difficult to enforce.

Question 25

A model platform solution has parallel requirements with separate owners. Workstream one must simplify authentication, serialization, and response typing in a Python service; workstream two must surface network timeouts differently from a completed response that declines a task. Which combination meets both?

  1. First component: Use the official SDK for standard requests. Second component: Use direct HTTP REST integration for unsupported runtimes.
  2. First component: Use the official SDK for standard requests. Second component: Separate transport failures from model refusals.
  3. First component: Use connection pooling for repeated HTTP calls. Second component: Separate transport failures from model refusals.
  4. First component: Use direct HTTP REST integration for unsupported runtimes. Second component: Use connection pooling for repeated HTTP calls.

Explanation

Correct Answer

B:

The first proposal fits its assigned role because maintained client libraries wrap the HTTP API and expose structured response objects, while the second proposal fits its assigned role because a transport error occurs before a valid application-level completion; refusals are model outcomes; their respective acceptance criteria are to simplify authentication, serialization, and response typing in a Python service and to surface network timeouts differently from a completed response that declines a task.

Incorrect Answers

A:

The first proposal fits its assigned role because maintained client libraries wrap the HTTP API and expose structured response objects, while the second proposal serves the different purpose to call the Messages endpoint from a language without a mature provider SDK instead of the assigned second role because a standards-based HTTP client can send JSON requests with the correct headers and decode responses; their respective acceptance criteria are to simplify authentication, serialization, and response typing in a Python service and to surface network timeouts differently from a completed response that declines a task.

C:

The first proposal serves the different purpose to reduce setup overhead when a service sends many short requests instead of the assigned first role because reusable clients avoid unnecessary connection establishment under compatible networking conditions, while the second proposal fits its assigned role because a transport error occurs before a valid application-level completion; refusals are model outcomes; their respective acceptance criteria are to simplify authentication, serialization, and response typing in a Python service and to surface network timeouts differently from a completed response that declines a task.

D:

The first proposal serves the different purpose to call the Messages endpoint from a language without a mature provider SDK instead of the assigned first role because a standards-based HTTP client can send JSON requests with the correct headers and decode responses, while the second proposal serves the different purpose to reduce setup overhead when a service sends many short requests instead of the assigned second role because reusable clients avoid unnecessary connection establishment under compatible networking conditions; their respective acceptance criteria are to simplify authentication, serialization, and response typing in a Python service and to surface network timeouts differently from a completed response that declines a task.

Question 26

A technical specification for a data pipeline application assigns two jobs to separate modules: module A must reject forged callbacks from an external asynchronous job service, and module B must keep an interrupted invoice-creation request from generating duplicate invoices. Which two-part solution is appropriate?

  1. First component: Validate webhook signatures. Second component: Avoid retrying non-idempotent writes blindly.
  2. First component: Use explicit HTTP method semantics. Second component: Use exponential backoff with jitter.
  3. First component: Validate webhook signatures. Second component: Use explicit HTTP method semantics.
  4. First component: Use exponential backoff with jitter. Second component: Avoid retrying non-idempotent writes blindly.

Explanation

Correct Answer

A:

The first proposal fits its assigned role because cryptographic verification checks the sender and payload integrity before processing the event, while the second proposal fits its assigned role because retries of side-effecting actions need idempotency keys or transaction reconciliation; their respective acceptance criteria are to reject forged callbacks from an external asynchronous job service and to keep an interrupted invoice-creation request from generating duplicate invoices.

Incorrect Answers

B:

The first proposal serves the different purpose to prevent a supposedly read-only status lookup from changing a server-side record instead of the assigned first role because gET should retrieve resources while state changes use appropriate mutation methods and authorization checks, while the second proposal serves the different purpose to recover from transient 429 responses without synchronized retry storms instead of the assigned second role because increasing randomized delays reduce contention and respect service throttling signals; their respective acceptance criteria are to reject forged callbacks from an external asynchronous job service and to keep an interrupted invoice-creation request from generating duplicate invoices.

C:

The first proposal fits its assigned role because cryptographic verification checks the sender and payload integrity before processing the event, while the second proposal serves the different purpose to prevent a supposedly read-only status lookup from changing a server-side record instead of the assigned second role because gET should retrieve resources while state changes use appropriate mutation methods and authorization checks; their respective acceptance criteria are to reject forged callbacks from an external asynchronous job service and to keep an interrupted invoice-creation request from generating duplicate invoices.

D:

The first proposal serves the different purpose to recover from transient 429 responses without synchronized retry storms instead of the assigned first role because increasing randomized delays reduce contention and respect service throttling signals, while the second proposal fits its assigned role because retries of side-effecting actions need idempotency keys or transaction reconciliation; their respective acceptance criteria are to reject forged callbacks from an external asynchronous job service and to keep an interrupted invoice-creation request from generating duplicate invoices.

Question 27

A latency optimization review requires two independent, explicitly defined outcomes: the first workstream must brainstorm diverse naming candidates without demanding identical wording on every run; a second must avoid confusing allowable prompt size with maximum generated response length. Which TWO actions directly provide these capabilities? Choose TWO.

  1. Use contrasting few-shot examples
  2. Use varied sampling only for ideation
  3. Distinguish input from output token limits
  4. Recognize next-token generation
  5. Use one representative example

Explanation

Correct Answers

B:

Greater sampling diversity can produce a wider range of suggestions for exploratory work; this directly supports task 1 to brainstorm diverse naming candidates without demanding identical wording on every run, not the separate task or tasks to avoid confusing allowable prompt size with maximum generated response length; it is not appropriate as the sole correctness control for structured production data.

C:

The context budget constrains total material considered while output ceilings bound what a single generation can emit; this directly supports task 2 to avoid confusing allowable prompt size with maximum generated response length, not the separate task or tasks to brainstorm diverse naming candidates without demanding identical wording on every run; raising an output cap does not make an oversized input fit.

Incorrect Answers

A:

Multiple representative cases help define a decision boundary and reduce ambiguity; this implements the different outcome to distinguish borderline policy labels with examples of both allowed and disallowed cases rather than the required tasks to avoid confusing allowable prompt size with maximum generated response length and to brainstorm diverse naming candidates without demanding identical wording on every run, and repeating near-identical samples does not cover edge cases, so it is not a direct solution to any of the assigned workstreams.

D:

The model predicts likely continuations conditioned on context rather than querying an authoritative truth database by default; this implements the different outcome to explain why a plausible answer can still contain an unsupported factual claim rather than the required tasks to avoid confusing allowable prompt size with maximum generated response length and to brainstorm diverse naming candidates without demanding identical wording on every run, and fluent prose is not evidence that a source was verified, so it is not a direct solution to any of the assigned workstreams.

E:

A single demonstration can anchor the expected shape and tone of a generated response; this implements the different outcome to clarify a specific formatting convention that a plain instruction was not conveying rather than the required tasks to avoid confusing allowable prompt size with maximum generated response length and to brainstorm diverse naming candidates without demanding identical wording on every run, and an example is not a formal guarantee that every response will conform, so it is not a direct solution to any of the assigned workstreams.

Question 28

The owner of a document processing integration wants to reject a malformed tool argument payload before calling a payroll service. Which proposal provides the required mechanism?

  1. Correlate distributed traces across services
  2. Separate integration tests from unit tests
  3. Avoid retrying non-idempotent writes blindly
  4. Keep secrets out of source control
  5. Validate JSON before deserialization

Explanation

Correct Answer

E:

A parser and schema validator catch missing types and invalid syntax before side effects; this behavior directly supports the need to reject a malformed tool argument payload before calling a payroll service. Model-generated text is not trustworthy merely because it looks structured.

Incorrect Answers

A:

A trace identifier links timing and errors through multiple process boundaries; this applies to the different goal to debug where a request stalls between the web UI, queue, and Claude gateway, not the requirement to reject a malformed tool argument payload before calling a payroll service; local logs with unrelated IDs cannot reconstruct the critical path reliably.

B:

When an application needs to verify that a real staging tool and its API contract work together, integration tests exercise boundaries and serialization that isolated mocks may miss; the current capability must instead reject a malformed tool argument payload before calling a payroll service, while unit tests can pass while authentication or payload mapping is broken.

C:

Retries of side-effecting actions need idempotency keys or transaction reconciliation; the supported outcome is to keep an interrupted invoice-creation request from generating duplicate invoices, whereas this scenario requires the ability to reject a malformed tool argument payload before calling a payroll service; network timeouts do not prove the first write failed.

D:

This approach serves the separate need to prevent API credentials from being exposed through shared repository history because secret stores or environment injection keep production tokens out of committed code; it does not establish the outcome to reject a malformed tool argument payload before calling a payroll service, and deleting a secret from the latest commit may not remove it from earlier revisions.

Question 29

Two separate tickets in an inference routing backlog require different controls. One needs to deliver tokens to a browser without maintaining a bidirectional command channel; the other needs to simplify authentication, serialization, and response typing in a Python service. Which implementation pair is correct?

  1. First component: Use connection pooling for repeated HTTP calls. Second component: Send the correct API authentication header.
  2. First component: Use a queue for burst absorption. Second component: Use connection pooling for repeated HTTP calls.
  3. First component: Choose SSE for one-way streamed responses. Second component: Use connection pooling for repeated HTTP calls.
  4. First component: Choose SSE for one-way streamed responses. Second component: Use the official SDK for standard requests.
  5. First component: Send the correct API authentication header. Second component: Use the official SDK for standard requests.

Explanation

Correct Answer

D:

The first proposal fits its assigned role because server-sent events provide ordered server-to-client updates over an HTTP stream, while the second proposal fits its assigned role because maintained client libraries wrap the HTTP API and expose structured response objects; their respective acceptance criteria are to deliver tokens to a browser without maintaining a bidirectional command channel and to simplify authentication, serialization, and response typing in a Python service.

Incorrect Answers

A:

The first proposal serves the different purpose to reduce setup overhead when a service sends many short requests instead of the assigned first role because reusable clients avoid unnecessary connection establishment under compatible networking conditions, while the second proposal serves the different purpose to avoid unauthenticated model requests from a backend worker instead of the assigned second role because the API validates the provided credential before accepting a request; their respective acceptance criteria are to deliver tokens to a browser without maintaining a bidirectional command channel and to simplify authentication, serialization, and response typing in a Python service.

B:

The first proposal serves the different purpose to smooth sudden ingestion spikes before workers call a capacity-limited model API instead of the assigned first role because queueing decouples producers from consumer throughput and enables controlled retries, while the second proposal serves the different purpose to reduce setup overhead when a service sends many short requests instead of the assigned second role because reusable clients avoid unnecessary connection establishment under compatible networking conditions; their respective acceptance criteria are to deliver tokens to a browser without maintaining a bidirectional command channel and to simplify authentication, serialization, and response typing in a Python service.

C:

The first proposal fits its assigned role because server-sent events provide ordered server-to-client updates over an HTTP stream, while the second proposal serves the different purpose to reduce setup overhead when a service sends many short requests instead of the assigned second role because reusable clients avoid unnecessary connection establishment under compatible networking conditions; their respective acceptance criteria are to deliver tokens to a browser without maintaining a bidirectional command channel and to simplify authentication, serialization, and response typing in a Python service.

E:

The first proposal serves the different purpose to avoid unauthenticated model requests from a backend worker instead of the assigned first role because the API validates the provided credential before accepting a request, while the second proposal fits its assigned role because maintained client libraries wrap the HTTP API and expose structured response objects; their respective acceptance criteria are to deliver tokens to a browser without maintaining a bidirectional command channel and to simplify authentication, serialization, and response typing in a Python service.

Question 30

A document processing review requires two independent, explicitly defined outcomes: the first workstream must debug where a request stalls between the web UI, queue, and Claude gateway; a second must recover from transient 429 responses without synchronized retry storms. Which TWO actions directly provide these capabilities? Choose TWO.

  1. Use explicit HTTP method semantics
  2. Use exponential backoff with jitter
  3. Correlate distributed traces across services
  4. Enforce concurrency limits
  5. Pin dependencies with lockfiles

Explanation

Correct Answers

B:

Increasing randomized delays reduce contention and respect service throttling signals; this directly supports task 2 to recover from transient 429 responses without synchronized retry storms, not the separate task or tasks to debug where a request stalls between the web UI, queue, and Claude gateway; immediate identical retries can extend an overload event.

C:

A trace identifier links timing and errors through multiple process boundaries; this directly supports task 1 to debug where a request stalls between the web UI, queue, and Claude gateway, not the separate task or tasks to recover from transient 429 responses without synchronized retry storms; local logs with unrelated IDs cannot reconstruct the critical path reliably.

Incorrect Answers

A:

GET should retrieve resources while state changes use appropriate mutation methods and authorization checks; this implements the different outcome to prevent a supposedly read-only status lookup from changing a server-side record rather than the required tasks to recover from transient 429 responses without synchronized retry storms and to debug where a request stalls between the web UI, queue, and Claude gateway, and putting write behavior behind a read endpoint undermines predictable retry and caching semantics, so it is not a direct solution to any of the assigned workstreams.

D:

A semaphore or bounded queue keeps in-flight work within measured capacity; this implements the different outcome to prevent bursts of document jobs from overwhelming rate limits and memory rather than the required tasks to recover from transient 429 responses without synchronized retry storms and to debug where a request stalls between the web UI, queue, and Claude gateway, and launching every request at once turns a throughput problem into throttling, so it is not a direct solution to any of the assigned workstreams.

E:

Locked versions and hashes constrain drift between build environments; this implements the different outcome to reproduce a working SDK installation across staging and production rather than the required tasks to recover from transient 429 responses without synchronized retry storms and to debug where a request stalls between the web UI, queue, and Claude gateway, and unbounded package updates can introduce new behavior during a routine rebuild, so it is not a direct solution to any of the assigned workstreams.

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