Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 44

Question 430

For a Python Claude API backend, the missing capability is to ensure refunds above a policy limit receive human sign-off. What should the developers implement?

  1. Apply content checks at suitable input and output boundaries.
  2. Use strict tool input validation for the exposed tool interface.
  3. Persist explicit job states with allowed transitions.
  4. Add application-owned approval logic around the refund tool.

Explanation

Correct Answer

D:

To ensure refunds above a policy limit receive human sign-off in this Python Claude API backend, decisions that create financial side effects require deterministic authorization checks beyond a model's recommendation; the proposed implementation directly supplies that capability.

Incorrect Answers

A:

The proposed alternative helps to filter risky user-visible content according to application policy because moderation is a defense layer applied around generation, not a substitute for authenticated tool access or scoped retrieval; that mechanism does not meet the different requirement to ensure refunds above a policy limit receive human sign-off.

B:

The proposed alternative helps to reject tool arguments that fail a declared shape before tool execution because a strict schema constrains the arguments a model may emit, but application authorization and business-rule checks remain necessary; that mechanism does not meet the different requirement to ensure refunds above a policy limit receive human sign-off.

C:

The proposed alternative helps to distinguish queued, running, waiting-for-approval, failed, and completed agent jobs because a state machine prevents an incomplete tool operation from being shown as successful and makes retries and cancellation predictable; that mechanism does not meet the different requirement to ensure refunds above a policy limit receive human sign-off.

Question 431

Two acceptance tests block a developer workstation running Claude Code: one requires the ability to continue a previously interrupted Claude Code work session; the other requires the ability to enforce a team-approved permission policy despite local developer preferences. Which plan clears both tests?

  1. Perform both steps: (1) Add concise project guidance to the appropriate CLAUDE.md file; (2) define policy in the appropriate managed or shared settings scope.
  2. Perform both steps: (1) Use the supported session continuation or resume mechanism; (2) define policy in the appropriate managed or shared settings scope.
  3. Perform both steps: (1) Use the supported session continuation or resume mechanism; (2) configure a specialized claude code subagent with a narrow remit.
  4. Perform both steps: (1) Add concise project guidance to the appropriate CLAUDE.md file; (2) configure a specialized claude code subagent with a narrow remit.

Explanation

Correct Answer

B:

The first change enables the team to continue a previously interrupted Claude Code work session because restoring session context reduces lost work while preserving continuity of prior tool activity; opening a new unrelated session does not do the same; the second addresses the separate need to enforce a team-approved permission policy despite local developer preferences because Claude Code has multiple configuration scopes; binding policy to the required scope matters more than placing a rule in an arbitrary JSON file.

Incorrect Answers

A:

The second step meets the requirement to enforce a team-approved permission policy despite local developer preferences because Claude Code has multiple configuration scopes; binding policy to the required scope matters more than placing a rule in an arbitrary JSON file; the first step would only help to give Claude Code repository-specific build and test instructions for future sessions because Claude Code reads persistent project instructions from its memory hierarchy; one-off chat messages do not reliably establish durable repository guidance, leaving the separate requirement to continue a previously interrupted Claude Code work session unmet.

C:

The first step is appropriate for the requirement to continue a previously interrupted Claude Code work session because restoring session context reduces lost work while preserving continuity of prior tool activity; opening a new unrelated session does not do the same; the other step instead aims to delegate an independent code review without filling the main coding session with details because subagents can focus on a task and return a summary, limiting noise in the parent session and allowing separate permissions when configured, which does not meet the second requirement to enforce a team-approved permission policy despite local developer preferences.

D:

Both steps target different results—give Claude Code repository-specific build and test instructions for future sessions and delegate an independent code review without filling the main coding session with details—because Claude Code reads persistent project instructions from its memory hierarchy; one-off chat messages do not reliably establish durable repository guidance and subagents can focus on a task and return a summary, limiting noise in the parent session and allowing separate permissions when configured; the required capabilities are instead to continue a previously interrupted Claude Code work session and to enforce a team-approved permission policy despite local developer preferences.

Question 432

A design review for a regression review for a tool-using service asks for support to explain why an agent loop is stuck after a model requested an external function, together with a way to find why a JSON record unexpectedly ends midway through a field. Select the two appropriate changes. Choose TWO.

  1. Trace the API credential and access permissions for the 401 or 403 response.
  2. Check that each tool_use identifier receives one matching tool_result.
  3. Inspect stop_reason and output limits before changing the parser.
  4. Compare versioned prompts and representative eval cases against model responses.
  5. Inspect the HTTP 429 status and rate-limit metadata before retrying.

Explanation

Correct Answers

B:

To explain why an agent loop is stuck after a model requested an external function as part of a design that must also find why a JSON record unexpectedly ends midway through a field, an unmatched tool call leaves a protocol-level continuation gap, even if the handler actually ran successfully; this is why the action is one of the two required selections.

C:

To find why a JSON record unexpectedly ends midway through a field as part of a design that must also explain why an agent loop is stuck after a model requested an external function, a max_tokens stop indicates incomplete generation; the parser may be working correctly against a response that was never finished; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to identify why a request is rejected before any inference happens because unauthorized access fails at the transport or authorization boundary, not because Claude misunderstood the prompt; the required outcomes are instead to explain why an agent loop is stuck after a model requested an external function and to find why a JSON record unexpectedly ends midway through a field.

D:

This alternative serves the separate goal to distinguish a syntactically valid but low-quality answer from a transport outage because a successful HTTP call only proves the protocol worked; quality regressions require output-level evidence and controlled comparisons; the required outcomes are instead to explain why an agent loop is stuck after a model requested an external function and to find why a JSON record unexpectedly ends midway through a field.

E:

This alternative serves the separate goal to distinguish an upstream throttling response from a malformed model answer because a rate limit is a request-layer capacity problem; the response needs controlled pacing rather than a prompt rewrite; the required outcomes are instead to explain why an agent loop is stuck after a model requested an external function and to find why a JSON record unexpectedly ends midway through a field.

Question 433

The technical owner of a business workflow that calls Claude has to need schema-conformant JSON for a downstream parser at generation time. A separate requirement is to keep a browser application from exposing the secret used for model calls. Which two-part remediation covers both?

  1. Perform both steps: (1) Configure supported structured outputs with a JSON Schema; (2) queue or reject excess requests using bounded admission control.
  2. Perform both steps: (1) Configure supported structured outputs with a JSON Schema; (2) route claude requests through a trusted backend service.
  3. Perform both steps: (1) Store and retrieve message history by authenticated tenant and conversation ID; (2) queue or reject excess requests using bounded admission control.
  4. Perform both steps: (1) Store and retrieve message history by authenticated tenant and conversation ID; (2) route claude requests through a trusted backend service.

Explanation

Correct Answer

B:

The first change enables the team to need schema-conformant JSON for a downstream parser at generation time because constrained JSON generation targets machine-parseable fields and types; free-form prompting alone does not establish the same guarantee; the second addresses the separate need to keep a browser application from exposing the secret used for model calls because a backend controls credentials and access policy while the UI sends user input through an authenticated application endpoint.

Incorrect Answers

A:

The first step is appropriate for the requirement to need schema-conformant JSON for a downstream parser at generation time because constrained JSON generation targets machine-parseable fields and types; free-form prompting alone does not establish the same guarantee; the other step instead aims to shield customer traffic when inference capacity is exhausted because controlled queuing protects system latency and prevents unbounded retries from overwhelming downstream provider quotas, which does not meet the second requirement to keep a browser application from exposing the secret used for model calls.

C:

Both steps target different results—prevent chat transcripts from different accounts entering the same Claude request and shield customer traffic when inference capacity is exhausted—because session hygiene requires server-enforced isolation; model instructions cannot repair a history assembled from the wrong user and controlled queuing protects system latency and prevents unbounded retries from overwhelming downstream provider quotas; the required capabilities are instead to need schema-conformant JSON for a downstream parser at generation time and to keep a browser application from exposing the secret used for model calls.

D:

The second step meets the requirement to keep a browser application from exposing the secret used for model calls because a backend controls credentials and access policy while the UI sends user input through an authenticated application endpoint; the first step would only help to prevent chat transcripts from different accounts entering the same Claude request because session hygiene requires server-enforced isolation; model instructions cannot repair a history assembled from the wrong user, leaving the separate requirement to need schema-conformant JSON for a downstream parser at generation time unmet.

Question 434

An end-to-end test for a token-consumption optimization sprint checks whether it can choose a cheaper model only when the workload-level economics improve and cut cost for a nightly noninteractive summarization workload. Which plan should be approved to satisfy both checks?

  1. Perform both steps: (1) Use more exploratory sampling where supported and measure quality; (2) prefer a maintained sdk when its typed methods and retry behavior fit the integration.
  2. Perform both steps: (1) Compare quality-adjusted cost per accepted result across candidate models; (2) evaluate message batches instead of synchronous requests.
  3. Perform both steps: (1) Use more exploratory sampling where supported and measure quality; (2) evaluate message batches instead of synchronous requests.
  4. Perform both steps: (1) Compare quality-adjusted cost per accepted result across candidate models; (2) prefer a maintained sdk when its typed methods and retry behavior fit the integration.

Explanation

Correct Answer

B:

The first change enables the team to choose a cheaper model only when the workload-level economics improve because a lower token price can be offset by retries, higher failure rates, or excessive output, so accepted-result cost is the useful measure; the second addresses the separate need to cut cost for a nightly noninteractive summarization workload because asynchronous batches can offer discounted processing when the application can tolerate deferred completion and result collection.

Incorrect Answers

A:

Both steps target different results—invite stylistic variety during noncritical brainstorming and choose between writing raw REST requests and using an official client library—because higher sampling randomness broadens candidate phrasing but can reduce predictability, so it suits ideation better than strict extraction and sDKs wrap the same underlying service API and can simplify serialization, error types, and streaming without removing protocol constraints; the required capabilities are instead to choose a cheaper model only when the workload-level economics improve and to cut cost for a nightly noninteractive summarization workload.

C:

The second step meets the requirement to cut cost for a nightly noninteractive summarization workload because asynchronous batches can offer discounted processing when the application can tolerate deferred completion and result collection; the first step would only help to invite stylistic variety during noncritical brainstorming because higher sampling randomness broadens candidate phrasing but can reduce predictability, so it suits ideation better than strict extraction, leaving the separate requirement to choose a cheaper model only when the workload-level economics improve unmet.

D:

The first step is appropriate for the requirement to choose a cheaper model only when the workload-level economics improve because a lower token price can be offset by retries, higher failure rates, or excessive output, so accepted-result cost is the useful measure; the other step instead aims to choose between writing raw REST requests and using an official client library because sDKs wrap the same underlying service API and can simplify serialization, error types, and streaming without removing protocol constraints, which does not meet the second requirement to cut cost for a nightly noninteractive summarization workload.

Question 435

A design review for a tool-using incident triage agent asks for support to continue a paused multi-step task with its legitimate prior context, together with a way to build a Claude-powered coding agent with an established runtime for tool orchestration. Select the two appropriate changes. Choose TWO.

  1. Catch tool failures and report typed errors into the agent loop.
  2. Use the Claude Agent SDK for the agent's execution harness.
  3. Persist and resume the agent session using supported state handling.
  4. Persist checkpoints for tool outputs and accepted decisions.
  5. Deploy the agent runner in the organization's own environment.

Explanation

Correct Answers

B:

To build a Claude-powered coding agent with an established runtime for tool orchestration as part of a design that must also continue a paused multi-step task with its legitimate prior context, the SDK provides structured primitives for agent sessions and tool execution, reducing the custom loop infrastructure the application must maintain; this is why the action is one of the two required selections.

C:

To continue a paused multi-step task with its legitimate prior context as part of a design that must also build a Claude-powered coding agent with an established runtime for tool orchestration, session continuity preserves relevant work across invocations; discarding session state can cause repeated tools or lost intermediate decisions; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to keep a failed tool from being mistaken for successful completion of an objective because explicit failure messages let the model or controller choose recovery; swallowing exceptions makes later actions rely on results that never existed; the required outcomes are instead to continue a paused multi-step task with its legitimate prior context and to build a Claude-powered coding agent with an established runtime for tool orchestration.

D:

This alternative serves the separate goal to recover a multi-hour agent workflow after an infrastructure interruption because durable checkpoints let the orchestrator resume from a known stage rather than rerunning costly or side-effecting actions; the required outcomes are instead to continue a paused multi-step task with its legitimate prior context and to build a Claude-powered coding agent with an established runtime for tool orchestration.

E:

This alternative serves the separate goal to keep orchestration runtime and internal tool handlers inside company-managed infrastructure because self-hosting puts tool execution, networking, and operational controls under the application's responsibility, subject to its own deployment constraints; the required outcomes are instead to continue a paused multi-step task with its legitimate prior context and to build a Claude-powered coding agent with an established runtime for tool orchestration.

Question 436

To ship a TypeScript Claude integration, the application must authenticate a server-to-server Claude API request and avoid retrying a malformed request that the API consistently rejects. Which pair of listed actions is necessary for those stated capabilities? Choose TWO.

  1. Use the API's documented POST request with a JSON body.
  2. Classify permanent client errors separately from transient failures.
  3. Place a valid secret in the documented API authentication header.
  4. Authenticate incoming callbacks and validate their signatures where supported.
  5. Supply the required API version header.

Explanation

Correct Answers

B:

To avoid retrying a malformed request that the API consistently rejects as part of a design that must also authenticate a server-to-server Claude API request, a deterministic validation or authorization error requires fixing the request or credential, not another identical attempt; this is why the action is one of the two required selections.

C:

To authenticate a server-to-server Claude API request as part of a design that must also avoid retrying a malformed request that the API consistently rejects, API authentication is a request transport concern; placing a key inside the prompt body neither authenticates the request nor protects it; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to send a model inference request with the operation expected by the REST endpoint because a state-changing inference operation is submitted as an HTTP request with an explicit method and payload, not inferred from a browser URL alone; the required outcomes are instead to authenticate a server-to-server Claude API request and to avoid retrying a malformed request that the API consistently rejects.

D:

This alternative serves the separate goal to trust a callback only after verifying who actually sent it because network arrival alone does not prove a webhook's origin; authenticity checks prevent forged events from changing agent state; the required outcomes are instead to authenticate a server-to-server Claude API request and to avoid retrying a malformed request that the API consistently rejects.

E:

This alternative serves the separate goal to make a direct REST request with a declared supported API contract because the version header selects the protocol contract supported by the endpoint, preventing accidental dependence on unspecified behavior; the required outcomes are instead to authenticate a server-to-server Claude API request and to avoid retrying a malformed request that the API consistently rejects.

Question 437

While planning a model selection trial, the product owner specifies that it must allow supported models to vary internal reasoning effort according to task difficulty. Which solution fulfills the requirement?

  1. Use more exploratory sampling where supported and measure quality.
  2. Tune sampling for lower variance while retaining output validation.
  3. Measure or count input and output tokens with the supported tokenizer or API.
  4. Configure supported adaptive thinking behavior rather than forcing a fixed pattern.

Explanation

Correct Answer

D:

To allow supported models to vary internal reasoning effort according to task difficulty in this model selection trial, adaptive reasoning responds to problem complexity, but supported modes differ by model and release and should not be assumed universal; the proposed implementation directly supplies that capability.

Incorrect Answers

A:

The proposed alternative helps to invite stylistic variety during noncritical brainstorming because higher sampling randomness broadens candidate phrasing but can reduce predictability, so it suits ideation better than strict extraction; that mechanism does not meet the different requirement to allow supported models to vary internal reasoning effort according to task difficulty.

B:

The proposed alternative helps to favor more consistent word choices in a constrained transformation task when supported because temperature influences token sampling; it cannot by itself guarantee truth, policy compliance, or reproducible structured output; that mechanism does not meet the different requirement to allow supported models to vary internal reasoning effort according to task difficulty.

C:

The proposed alternative helps to estimate model usage from encoded text rather than counting words because billing and context consumption track tokens, which do not map one-to-one to English words or characters; that mechanism does not meet the different requirement to allow supported models to vary internal reasoning effort according to task difficulty.

Question 438

An operational review of an agent access-control assessment reports distinct failures: it cannot reject a malicious instruction inserted into a web-search result rather than the user's direct request, and it cannot stop a model-requested webhook from sending sensitive case text to an unapproved host. Which two repairs are required? Choose TWO.

  1. Isolate external search results and validate any resulting tool action independently.
  2. Enforce outbound destination allowlists at the tool execution layer.
  3. Define fail-closed behavior for an enforcement hook and test failure paths.
  4. Record authenticated approval events with actor, scope, and request correlation.
  5. Use layered policy enforcement and restrict powerful tools independently of chat text.

Explanation

Correct Answers

A:

To reject a malicious instruction inserted into a web-search result rather than the user's direct request as part of a design that must also stop a model-requested webhook from sending sensitive case text to an unapproved host, indirect attacks exploit lower-trust retrieval surfaces, so trusting search snippets as operational directives crosses a security boundary; this is why the action is one of the two required selections.

B:

To stop a model-requested webhook from sending sensitive case text to an unapproved host as part of a design that must also reject a malicious instruction inserted into a web-search result rather than the user's direct request, network restrictions prevent data egress even if injected instructions convince a model to request the call; this is why the action is one of the two required selections.

Incorrect Answers

C:

This alternative serves the separate goal to avoid accidentally allowing critical writes when a safety check crashes because a guardrail that silently skips checks on errors is not an effective boundary; failures must prevent the protected action when policy requires it; the required outcomes are instead to reject a malicious instruction inserted into a web-search result rather than the user's direct request and to stop a model-requested webhook from sending sensitive case text to an unapproved host.

D:

This alternative serves the separate goal to demonstrate who approved a privileged tool action and when because an audit trail binds authorization to a specific action, making privilege use reviewable without treating model-generated text as proof of consent; the required outcomes are instead to reject a malicious instruction inserted into a web-search result rather than the user's direct request and to stop a model-requested webhook from sending sensitive case text to an unapproved host.

E:

This alternative serves the separate goal to prevent an adversarial user from bypassing safety policy by roleplay instructions because adversarial prompting can target instruction hierarchy, so product controls and tool authorization must remain effective when wording is manipulated; the required outcomes are instead to reject a malicious instruction inserted into a web-search result rather than the user's direct request and to stop a model-requested webhook from sending sensitive case text to an unapproved host.

Question 439

In a customer-facing AI feature, testing reveals that the system cannot state how much interruption a customer-facing AI feature can tolerate and cannot identify which search services and internal systems are critical to the workflow. Select the two targeted actions that close these gaps. Choose TWO.

  1. Refactor in bounded steps under existing tests.
  2. Define measurable answer-format and citation acceptance criteria.
  3. Define availability objectives with failure and recovery criteria.
  4. List upstream systems with permissions, quotas, and failure behavior.
  5. Track latency, failure rates, token use, and output-quality signals over time.

Explanation

Correct Answers

C:

To state how much interruption a customer-facing AI feature can tolerate as part of a design that must also identify which search services and internal systems are critical to the workflow, reliability requirements determine fallback, retries, health monitoring, and degraded behavior; they are distinct from the response's linguistic quality; this is why the action is one of the two required selections.

D:

To identify which search services and internal systems are critical to the workflow as part of a design that must also state how much interruption a customer-facing AI feature can tolerate, a dependency inventory exposes the operational and authorization constraints that a purely model-centered design misses; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to preserve behavior while replacing a fragile parser in an AI integration because small changes constrain regression scope; a wholesale rewrite makes it harder to isolate whether model or parsing behavior changed; the required outcomes are instead to state how much interruption a customer-facing AI feature can tolerate and to identify which search services and internal systems are critical to the workflow.

B:

This alternative serves the separate goal to translate a request for citation-backed answers into observable application behavior because a functional requirement specifies what the user must receive and how success is evaluated, not merely that the team will use a language model; the required outcomes are instead to state how much interruption a customer-facing AI feature can tolerate and to identify which search services and internal systems are critical to the workflow.

E:

This alternative serves the separate goal to detect rising inference errors after a feature reaches production because production health spans infrastructure and model behavior, so a healthy HTTP server alone does not prove the AI feature is functioning; the required outcomes are instead to state how much interruption a customer-facing AI feature can tolerate and to identify which search services and internal systems are critical to the workflow.

Question 440

A production readiness review of a token-consumption optimization sprint finds that the team cannot yet investigate inconsistent throughput during a traffic spike or keep a server responsive while several independent model calls wait on network I/O. Which paired changes directly remedy the two gaps?

  1. Perform both steps: (1) Design an approved fallback model path with compatibility checks; (2) use asynchronous client methods and limit in-flight calls.
  2. Perform both steps: (1) Record rate-limit status and relevant response headers alongside request metrics; (2) include several carefully selected and diverse demonstrations.
  3. Perform both steps: (1) Record rate-limit status and relevant response headers alongside request metrics; (2) use asynchronous client methods and limit in-flight calls.
  4. Perform both steps: (1) Design an approved fallback model path with compatibility checks; (2) include several carefully selected and diverse demonstrations.

Explanation

Correct Answer

C:

The first change enables the team to investigate inconsistent throughput during a traffic spike because quota-related measurements explain capacity failures that may otherwise look like randomly slow model inference; the second addresses the separate need to keep a server responsive while several independent model calls wait on network I/O because event-loop based I/O overlaps waits, but rate and resource bounds are still necessary to avoid overload.

Incorrect Answers

A:

The second step meets the requirement to keep a server responsive while several independent model calls wait on network I/O because event-loop based I/O overlaps waits, but rate and resource bounds are still necessary to avoid overload; the first step would only help to preserve critical service during a temporary primary-model problem because fallbacks improve resilience only when downstream schemas, cost limits, and policy controls work with the alternate model, leaving the separate requirement to investigate inconsistent throughput during a traffic spike unmet.

B:

The first step is appropriate for the requirement to investigate inconsistent throughput during a traffic spike because quota-related measurements explain capacity failures that may otherwise look like randomly slow model inference; the other step instead aims to reduce inconsistent labeling across nuanced examples because multiple examples communicate decision boundaries and desired formatting, although irrelevant or misleading demonstrations can bias the model, which does not meet the second requirement to keep a server responsive while several independent model calls wait on network I/O.

D:

Both steps target different results—preserve critical service during a temporary primary-model problem and reduce inconsistent labeling across nuanced examples—because fallbacks improve resilience only when downstream schemas, cost limits, and policy controls work with the alternate model and multiple examples communicate decision boundaries and desired formatting, although irrelevant or misleading demonstrations can bias the model; the required capabilities are instead to investigate inconsistent throughput during a traffic spike and to keep a server responsive while several independent model calls wait on network I/O.

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