Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 18

Question 170

A model platform team owns two independent components. The first must reduce variation when a service maps documents into consistent categories; the second must avoid confusing allowable prompt size with maximum generated response length. Which paired plan fulfills both?

  1. First component: Use low-temperature generation for repeatable extraction. Second component: Distinguish input from output token limits.
  2. First component: Recognize next-token generation. Second component: Distinguish input from output token limits.
  3. First component: Use low-temperature generation for repeatable extraction. Second component: Reduce reasoning effort on trivial routing.
  4. First component: Reduce reasoning effort on trivial routing. Second component: Recognize next-token generation.

Explanation

Correct Answer

A:

The first proposal fits its assigned role because lower randomness often improves consistency for narrow, constrained tasks, while the second proposal fits its assigned role because the context budget constrains total material considered while output ceilings bound what a single generation can emit; their respective acceptance criteria are to reduce variation when a service maps documents into consistent categories and to avoid confusing allowable prompt size with maximum generated response length.

Incorrect Answers

B:

The first proposal serves the different purpose to explain why a plausible answer can still contain an unsupported factual claim instead of the assigned first role because the model predicts likely continuations conditioned on context rather than querying an authoritative truth database by default, while the second proposal fits its assigned role because the context budget constrains total material considered while output ceilings bound what a single generation can emit; their respective acceptance criteria are to reduce variation when a service maps documents into consistent categories and to avoid confusing allowable prompt size with maximum generated response length.

C:

The first proposal fits its assigned role because lower randomness often improves consistency for narrow, constrained tasks, while the second proposal serves the different purpose to shorten latency and spend on tasks that need only basic categorization instead of the assigned second role because lower effort can suit predictable work where deeper reasoning adds little measurable quality; their respective acceptance criteria are to reduce variation when a service maps documents into consistent categories and to avoid confusing allowable prompt size with maximum generated response length.

D:

The first proposal serves the different purpose to shorten latency and spend on tasks that need only basic categorization instead of the assigned first role because lower effort can suit predictable work where deeper reasoning adds little measurable quality, while the second proposal serves the different purpose to explain why a plausible answer can still contain an unsupported factual claim instead of the assigned second role because the model predicts likely continuations conditioned on context rather than querying an authoritative truth database by default; their respective acceptance criteria are to reduce variation when a service maps documents into consistent categories and to avoid confusing allowable prompt size with maximum generated response length.

Question 171

A technical lead requests a solution that can contain cumulative token and tool cost across a recursive research session within a token economics project. Which design should be chosen?

  1. Set maximum step counts for agents
  2. Cache a stable shared prefix
  3. Track input and output usage separately
  4. Estimate output length before scaling

Explanation

Correct Answer

A:

A bound on iterations prevents repeated generation and calls from growing without limit; this behavior directly supports the need to contain cumulative token and tool cost across a recursive research session. A per-message output cap alone does not control total multi-step spend.

Incorrect Answers

B:

Prefix caching can lower repeated-input cost when the eligible prefix and cache conditions match; despite being useful when a service must reduce repeated processing of identical tool definitions and system instructions, it is not a solution to contain cumulative token and tool cost across a recursive research session; caching does not turn every changed prompt into an automatic hit.

C:

For the alternate objective to forecast cost for an application that sends long documents but returns brief classifications, metering by token category reveals the main cost driver and supports realistic forecasts; the feature here must contain cumulative token and tool cost across a recursive research session, and counting only request totals hides prompt-size variability.

D:

Cost models should include the generated token volume and task-specific variance; this applies to the different goal to avoid underbudgeting a report-writing feature whose completions are much longer than inputs, not the requirement to contain cumulative token and tool cost across a recursive research session; input-only estimates miss potentially dominant completion expense.

Question 172

Two features of a prompt engineering system are under review. One feature must prevent a customer message from overriding the application's escalation rules; the other must stop a summarizer from producing unrelated analysis instead of required action items. Which plan matches the respective needs?

  1. First component: Separate system policy from user content. Second component: Constrain response audience and purpose.
  2. First component: Specify the handling of missing information. Second component: State one measurable outcome per instruction.
  3. First component: Separate system policy from user content. Second component: State one measurable outcome per instruction.
  4. First component: Constrain response audience and purpose. Second component: Specify the handling of missing information.

Explanation

Correct Answer

C:

The first proposal fits its assigned role because stable higher-priority instructions should be represented in the designated instruction channel, while the second proposal fits its assigned role because clear task goals and acceptance criteria reduce competing interpretations; their respective acceptance criteria are to prevent a customer message from overriding the application's escalation rules and to stop a summarizer from producing unrelated analysis instead of required action items.

Incorrect Answers

A:

The first proposal fits its assigned role because stable higher-priority instructions should be represented in the designated instruction channel, while the second proposal serves the different purpose to generate a technical incident summary suitable for an on-call engineer instead of an executive instead of the assigned second role because audience constraints determine useful terminology, specificity, and level of detail; their respective acceptance criteria are to prevent a customer message from overriding the application's escalation rules and to stop a summarizer from producing unrelated analysis instead of required action items.

B:

The first proposal serves the different purpose to avoid invented contract dates when the uploaded document lacks them instead of the assigned first role because an explicit unknown or null rule makes uncertainty visible in the requested output, while the second proposal fits its assigned role because clear task goals and acceptance criteria reduce competing interpretations; their respective acceptance criteria are to prevent a customer message from overriding the application's escalation rules and to stop a summarizer from producing unrelated analysis instead of required action items.

D:

The first proposal serves the different purpose to generate a technical incident summary suitable for an on-call engineer instead of an executive instead of the assigned first role because audience constraints determine useful terminology, specificity, and level of detail, while the second proposal serves the different purpose to avoid invented contract dates when the uploaded document lacks them instead of the assigned second role because an explicit unknown or null rule makes uncertainty visible in the requested output; their respective acceptance criteria are to prevent a customer message from overriding the application's escalation rules and to stop a summarizer from producing unrelated analysis instead of required action items.

Question 173

A model platform team owns two independent components. The first must handle a rare complex architecture review where reasoning accuracy outweighs cost and latency; the second must classify many short messages where low latency and unit cost dominate. Which paired plan fulfills both?

  1. First component: Prefer a Sonnet-class model for balanced workloads. Second component: Prefer a Haiku-class model for lightweight routing.
  2. First component: Prefer a high-capability Opus-class model. Second component: Prefer a Haiku-class model for lightweight routing.
  3. First component: Pin model versions during a migration. Second component: Prefer a Sonnet-class model for balanced workloads.
  4. First component: Prefer a high-capability Opus-class model. Second component: Pin model versions during a migration.
  5. First component: Run task-specific model evaluations. Second component: Pin model versions during a migration.

Explanation

Correct Answer

B:

The first proposal fits its assigned role because the capability-oriented tier is suited to the hardest tasks after evaluation confirms an advantage, while the second proposal fits its assigned role because the lighter tier suits inexpensive, high-throughput tasks with modest reasoning demand; their respective acceptance criteria are to handle a rare complex architecture review where reasoning accuracy outweighs cost and latency and to classify many short messages where low latency and unit cost dominate.

Incorrect Answers

A:

The first proposal serves the different purpose to serve a mixed coding and support assistant at moderate latency and cost instead of the assigned first role because the balanced tier often offers a practical quality-throughput compromise, subject to task-level testing, while the second proposal fits its assigned role because the lighter tier suits inexpensive, high-throughput tasks with modest reasoning demand; their respective acceptance criteria are to handle a rare complex architecture review where reasoning accuracy outweighs cost and latency and to classify many short messages where low latency and unit cost dominate.

C:

The first proposal serves the different purpose to detect behavior changes before a newly released model reaches customers instead of the assigned first role because explicit IDs and a staged rollout make differences attributable and reversible, while the second proposal serves the different purpose to serve a mixed coding and support assistant at moderate latency and cost instead of the assigned second role because the balanced tier often offers a practical quality-throughput compromise, subject to task-level testing; their respective acceptance criteria are to handle a rare complex architecture review where reasoning accuracy outweighs cost and latency and to classify many short messages where low latency and unit cost dominate.

D:

The first proposal fits its assigned role because the capability-oriented tier is suited to the hardest tasks after evaluation confirms an advantage, while the second proposal serves the different purpose to detect behavior changes before a newly released model reaches customers instead of the assigned second role because explicit IDs and a staged rollout make differences attributable and reversible; their respective acceptance criteria are to handle a rare complex architecture review where reasoning accuracy outweighs cost and latency and to classify many short messages where low latency and unit cost dominate.

E:

The first proposal serves the different purpose to choose between releases using an actual organization's error and latency measurements instead of the assigned first role because representative tests expose differences hidden by generic capability labels, while the second proposal serves the different purpose to detect behavior changes before a newly released model reaches customers instead of the assigned second role because explicit IDs and a staged rollout make differences attributable and reversible; their respective acceptance criteria are to handle a rare complex architecture review where reasoning accuracy outweighs cost and latency and to classify many short messages where low latency and unit cost dominate.

Question 174

A production-readiness check for a Claude application component requires the service to require manual signoff when an AI-produced recommendation exceeds a financial threshold. Which approach passes this check?

  1. Identify authoritative source systems
  2. Document functional acceptance criteria
  3. Specify human-review triggers
  4. Calculate expected traffic and concurrency

Explanation

Correct Answer

C:

Decision thresholds turn governance expectations into testable escalation rules; this behavior directly supports the need to require manual signoff when an AI-produced recommendation exceeds a financial threshold. A generic instruction to use caution cannot enforce a payment approval.

Incorrect Answers

A:

For the alternate objective to prevent a Claude assistant from inventing account balances when it lacks current ledger data, source-of-truth mapping identifies which verified systems must supply facts and what freshness guarantees apply; the feature here must require manual signoff when an AI-produced recommendation exceeds a financial threshold, and model inference is not a substitute for a transactional database.

B:

Acceptance criteria specify inputs, outputs, and successful behavior that can be tested before release; this applies to the different goal to turn a vague request for an AI claims assistant into observable user actions and outputs, not the requirement to require manual signoff when an AI-produced recommendation exceeds a financial threshold; choosing a model first does not define what the application must accomplish.

D:

Workload assumptions drive rate-limit planning, queue capacity, and retry strategy; the supported outcome is to provision an interactive application for peak seasonal traffic, whereas this scenario requires the ability to require manual signoff when an AI-produced recommendation exceeds a financial threshold; average request volume alone hides burst behavior.

Question 175

An architecture review of a conversation design product lists two distinct outcomes: the first subsystem must save tokens when several retrieval chunks contain the same repeated policy section, and a second subsystem must let an assistant continue a long incident investigation without losing decisions. Which pair delivers these outcomes?

  1. First component: Use scoped subagent contexts. Second component: Separate long-term memory from active scratchpad.
  2. First component: Separate long-term memory from active scratchpad. Second component: Compact completed task history.
  3. First component: Deduplicate overlapping passages. Second component: Compact completed task history.
  4. First component: Deduplicate overlapping passages. Second component: Use scoped subagent contexts.

Explanation

Correct Answer

C:

The first proposal fits its assigned role because overlap detection reduces redundant context without throwing away unique clauses, while the second proposal fits its assigned role because a structured summary preserves established facts, pending work, and unresolved assumptions in fewer tokens; their respective acceptance criteria are to save tokens when several retrieval chunks contain the same repeated policy section and to let an assistant continue a long incident investigation without losing decisions.

Incorrect Answers

A:

The first proposal serves the different purpose to prevent a specialist's exhaustive document scan from contaminating the coordinator's working context instead of the assigned first role because isolated workers return concise findings while retaining intermediate detail within their own window, while the second proposal serves the different purpose to prevent outdated user preferences from being treated as current task instructions instead of the assigned second role because explicit storage and freshness rules govern what is retained and when it is reintroduced; their respective acceptance criteria are to save tokens when several retrieval chunks contain the same repeated policy section and to let an assistant continue a long incident investigation without losing decisions.

B:

The first proposal serves the different purpose to prevent outdated user preferences from being treated as current task instructions instead of the assigned first role because explicit storage and freshness rules govern what is retained and when it is reintroduced, while the second proposal fits its assigned role because a structured summary preserves established facts, pending work, and unresolved assumptions in fewer tokens; their respective acceptance criteria are to save tokens when several retrieval chunks contain the same repeated policy section and to let an assistant continue a long incident investigation without losing decisions.

D:

The first proposal fits its assigned role because overlap detection reduces redundant context without throwing away unique clauses, while the second proposal serves the different purpose to prevent a specialist's exhaustive document scan from contaminating the coordinator's working context instead of the assigned second role because isolated workers return concise findings while retaining intermediate detail within their own window; their respective acceptance criteria are to save tokens when several retrieval chunks contain the same repeated policy section and to let an assistant continue a long incident investigation without losing decisions.

Question 176

A production API review requires two independent, explicitly defined outcomes: the first workstream must avoid parsing safety refusals as valid JSON business records; a second must analyze a photographed shipping label with the supported multimodal Messages input. Which TWO actions directly provide these capabilities? Choose TWO.

  1. Handle the max_tokens stop_reason
  2. Handle the tool_use stop_reason
  3. Handle a refusal stop_reason safely
  4. Include an image content block
  5. Call the Messages API with explicit model and max_tokens

Explanation

Correct Answers

C:

A refusal is a deliberate non-completion path that requires a defined application response; this directly supports task 1 to avoid parsing safety refusals as valid JSON business records, not the separate task or tasks to analyze a photographed shipping label with the supported multimodal Messages input; repeatedly increasing the token limit does not resolve a content refusal.

D:

Image inputs are supplied as typed content with appropriate source data or references; this directly supports task 2 to analyze a photographed shipping label with the supported multimodal Messages input, not the separate task or tasks to avoid parsing safety refusals as valid JSON business records; describing a local image pathname in ordinary text does not transmit its pixels.

Incorrect Answers

A:

The finish reason identifies truncation, so a caller may need a continuation or a revised token budget; this implements the different outcome to recover a response that ended because its output budget was exhausted rather than the required tasks to analyze a photographed shipping label with the supported multimodal Messages input and to avoid parsing safety refusals as valid JSON business records, and treating that response as a complete answer can hide missing fields, so it is not a direct solution to any of the assigned workstreams.

B:

This stop reason means tool calls need to be executed and their results returned before synthesis; this implements the different outcome to continue the agent loop when Claude is waiting for external observations rather than the required tasks to analyze a photographed shipping label with the supported multimodal Messages input and to avoid parsing safety refusals as valid JSON business records, and terminating at this point strands an unfinished task, so it is not a direct solution to any of the assigned workstreams.

E:

A message request identifies the target model, a message list, and an output-token ceiling; this implements the different outcome to return an answer to a short user request through a backend service rather than the required tasks to analyze a photographed shipping label with the supported multimodal Messages input and to avoid parsing safety refusals as valid JSON business records, and the API is not a stateful chat session unless prior turns are supplied, so it is not a direct solution to any of the assigned workstreams.

Question 177

A language-model integration application is approaching release, but the team still needs to avoid underbudgeting a report-writing feature whose completions are much longer than inputs. What is the most appropriate implementation?

  1. Estimate output length before scaling
  2. Measure cache read tokens
  3. Keep volatile fields after reusable context
  4. Set budget alerts and tenant caps

Explanation

Correct Answer

A:

Cost models should include the generated token volume and task-specific variance; this behavior directly supports the need to avoid underbudgeting a report-writing feature whose completions are much longer than inputs. Input-only estimates miss potentially dominant completion expense.

Incorrect Answers

B:

Usage fields show how much eligible cached input was read rather than recomputed; the supported outcome is to verify that an optimization actually reuses an earlier prompt prefix, whereas this scenario requires the ability to avoid underbudgeting a report-writing feature whose completions are much longer than inputs; assuming identical UI prompts means a cache hit can conceal changed metadata or order.

C:

This approach serves the separate need to prevent request timestamps from invalidating a costly static prefix because stable instruction and reference sections should precede changing per-request details for effective prefix reuse; it does not establish the outcome to avoid underbudgeting a report-writing feature whose completions are much longer than inputs, and placing a changing value before the cache boundary undermines matching.

D:

Accounting plus enforced limits bounds exposure to unexpected traffic and agent loops; despite being useful when a service must stop a single customer from consuming the entire API allowance during a runaway workflow, it is not a solution to avoid underbudgeting a report-writing feature whose completions are much longer than inputs; usage dashboards alone only report overspend after the fact.

Question 178

A design review identifies a missing capability in a tool integration project: it must automate the same classify-review-export path across predictable submissions. Which change meets that goal?

  1. Use system or repository instructions for stable behavior
  2. Use subagents for independent expertise
  3. Combine tool access with human confirmation
  4. Use a deterministic workflow for fixed stages
  5. Publish an MCP server for reusable integrations

Explanation

Correct Answer

D:

Explicit application control makes each transition testable and easier to audit; this behavior directly supports the need to automate the same classify-review-export path across predictable submissions. An autonomous agent is unnecessary when the path is already fixed.

Incorrect Answers

A:

Instructions shape model behavior and task framing without executing external operations; despite being useful when a service must make a coding or API assistant follow organization-wide conventions, it is not a solution to automate the same classify-review-export path across predictable submissions; descriptive rules do not replace programmatic validation or IAM.

B:

For the alternate objective to delegate a security analysis while keeping the main coordinator's context concise, specialist contexts reduce tool-output contamination and support focused parallel work; the feature here must automate the same classify-review-export path across predictable submissions, and a reusable Skill alone does not create an independent worker with separate state.

C:

Tools supply actionable capabilities while a separate approval step governs sensitive execution; this applies to the different goal to enable a customer service agent to prepare but not autonomously approve account closures, not the requirement to automate the same classify-review-export path across predictable submissions; access to a function does not imply authority to invoke it without review.

E:

MCP standardizes tool and resource discovery across compatible applications; the supported outcome is to make one internal ticketing connector available to several agent clients, whereas this scenario requires the ability to automate the same classify-review-export path across predictable submissions; duplicated per-client wrappers increase maintenance across a growing ecosystem.

Question 179

During a privacy controls rollout, two different failure paths need remediation. The first component must let a support assistant read shipment status without issuing refunds; separately, the second must stop an unauthenticated user from invoking a dangerous account-management tool. Which changes should be adopted?

  1. First component: Enforce least-privilege tool permissions. Second component: Run adversarial prompt-injection tests.
  2. First component: Enforce least-privilege tool permissions. Second component: Require verified identity for privileged actions.
  3. First component: Run adversarial prompt-injection tests. Second component: Sandbox generated code execution.
  4. First component: Sandbox generated code execution. Second component: Require verified identity for privileged actions.

Explanation

Correct Answer

B:

The first proposal fits its assigned role because restricted scopes and read-only interfaces limit impact if a tool request is manipulated, while the second proposal fits its assigned role because authentication establishes the caller before authorization policies consider requested permissions; their respective acceptance criteria are to let a support assistant read shipment status without issuing refunds and to stop an unauthenticated user from invoking a dangerous account-management tool.

Incorrect Answers

A:

The first proposal fits its assigned role because restricted scopes and read-only interfaces limit impact if a tool request is manipulated, while the second proposal serves the different purpose to find vulnerabilities in a retrieval-augmented assistant before it accesses production tools instead of the assigned second role because purposeful hostile test cases probe whether the trust boundary holds under manipulated evidence; their respective acceptance criteria are to let a support assistant read shipment status without issuing refunds and to stop an unauthenticated user from invoking a dangerous account-management tool.

C:

The first proposal serves the different purpose to find vulnerabilities in a retrieval-augmented assistant before it accesses production tools instead of the assigned first role because purposeful hostile test cases probe whether the trust boundary holds under manipulated evidence, while the second proposal serves the different purpose to inspect proposed scripts without granting them production host privileges instead of the assigned second role because resource, filesystem, and network isolation contain failures and malicious effects; their respective acceptance criteria are to let a support assistant read shipment status without issuing refunds and to stop an unauthenticated user from invoking a dangerous account-management tool.

D:

The first proposal serves the different purpose to inspect proposed scripts without granting them production host privileges instead of the assigned first role because resource, filesystem, and network isolation contain failures and malicious effects, while the second proposal fits its assigned role because authentication establishes the caller before authorization policies consider requested permissions; their respective acceptance criteria are to let a support assistant read shipment status without issuing refunds and to stop an unauthenticated user from invoking a dangerous account-management tool.

Question 180

Two features of a customer platform system are under review. One feature must debug where a request stalls between the web UI, queue, and Claude gateway; the other must serve independent API calls without blocking a worker thread for each network wait. Which plan matches the respective needs?

  1. First component: Apply dependency injection for external clients. Second component: Use exponential backoff with jitter.
  2. First component: Use exponential backoff with jitter. Second component: Use asynchronous I/O for concurrent requests.
  3. First component: Correlate distributed traces across services. Second component: Use asynchronous I/O for concurrent requests.
  4. First component: Correlate distributed traces across services. Second component: Apply dependency injection for external clients.

Explanation

Correct Answer

C:

The first proposal fits its assigned role because a trace identifier links timing and errors through multiple process boundaries, while the second proposal fits its assigned role because async execution overlaps I/O-bound tasks while maintaining control over concurrency; their respective acceptance criteria are to debug where a request stalls between the web UI, queue, and Claude gateway and to serve independent API calls without blocking a worker thread for each network wait.

Incorrect Answers

A:

The first proposal serves the different purpose to test application logic without issuing real model calls or payments instead of the assigned first role because injected interfaces allow deterministic stubs and reduce coupling to vendors, 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 debug where a request stalls between the web UI, queue, and Claude gateway and to serve independent API calls without blocking a worker thread for each network wait.

B:

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 async execution overlaps I/O-bound tasks while maintaining control over concurrency; their respective acceptance criteria are to debug where a request stalls between the web UI, queue, and Claude gateway and to serve independent API calls without blocking a worker thread for each network wait.

D:

The first proposal fits its assigned role because a trace identifier links timing and errors through multiple process boundaries, while the second proposal serves the different purpose to test application logic without issuing real model calls or payments instead of the assigned second role because injected interfaces allow deterministic stubs and reduce coupling to vendors; their respective acceptance criteria are to debug where a request stalls between the web UI, queue, and Claude gateway and to serve independent API calls without blocking a worker thread for each network wait.

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