Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 26

Question 250

Finance needs to estimate inference spend for a particular model across mixed request lengths. Which production safeguard is most relevant?

  1. Alert on unusual token spikes
  2. Aggregate usage across multi-step tasks
  3. Compute cost from input and output token usage
  4. Include cache reads and writes in cost calculations

Explanation

Correct Answer

C:

Per-model input and output token rates combine with measured usage for accurate costing, subject to this requirement: number of messages alone is not an adequate unit of resource consumption.

Incorrect Answers

A:

Cost monitoring can detect and halt anomalous demand before it causes significant overspend; this cannot substitute for the behavior that per-model input and output token rates combine with measured usage for accurate costing, because monthly billing review alone detects runaway costs too late.

B:

Summing usage across the whole trace shows the true cost of satisfying one business request; in contrast, per-model input and output token rates combine with measured usage for accurate costing, while only counting the final call omits earlier orchestration expenses.

D:

Cache creation and cache retrieval have distinct billing treatment that affects economics; here, the controlling capability is that per-model input and output token rates combine with measured usage for accurate costing, but treating every token as an uncached input misstates true spend.

Question 251

A credential appears in a public build log and there is uncertainty about who copied it. Which measure directly addresses the problem?

  1. Monitor key usage anomalies
  2. Rotate a key after suspected leakage
  3. Check end-user permissions before using a customer tool
  4. Use separate credentials for development and production

Explanation

Correct Answer

B:

Revoking and replacing the affected secret reduces the time attackers can continue using it; its operational boundary is that hiding the log page alone cannot guarantee the secret was not captured.

Incorrect Answers

A:

Usage attribution, budget alerts, and access logs can expose unauthorized patterns for investigation; this cannot substitute for the behavior that revoking and replacing the affected secret reduces the time attackers can continue using it, because a key that still works is not evidence that its use is legitimate.

C:

The backend verifies the caller is allowed to access that specific object; this addresses a separate concern, not the fact that revoking and replacing the affected secret reduces the time attackers can continue using it; a valid session establishes identity but not ownership of every possible resource.

D:

Environment isolation makes accidental or compromised test execution less damaging; here, the controlling capability is that revoking and replacing the affected secret reduces the time attackers can continue using it, but a single unrestricted key creates unnecessary blast radius across environments.

Question 252

Every API call repeats a large policy document with duplicate paragraphs that add no relevant decisions. Which option best fits the described workflow?

  1. Cache a reusable prompt prefix
  2. Shorten redundant system context
  3. Count the complete request against context limits
  4. Use batch processing for noninteractive volume

Explanation

Correct Answer

B:

Trimming repeated instructions reduces input tokens without sacrificing unique requirements; cutting security-critical constraints to save tokens can be counterproductive.

Incorrect Answers

A:

Eligible prompt caching can avoid repeatedly processing unchanged prefix tokens; by comparison, the intended solution relies on this: trimming repeated instructions reduces input tokens without sacrificing unique requirements; cache savings require actual hits and depend on write/read prices.

C:

The context budget encompasses the relevant input contents and model-specific output allowances, whereas the requirement calls for another behavior: trimming repeated instructions reduces input tokens without sacrificing unique requirements; counting only the latest user text understates true token pressure.

D:

A batch interface is suited to many independent requests with no immediate response requirement; however, the critical mechanism here is that trimming repeated instructions reduces input tokens without sacrificing unique requirements, and interactive users cannot depend on eventual batch completion.

Question 253

A client receives a Claude response containing a tool_use block for a permitted inventory lookup. Which of the following would work as specified?

  1. Return a tool_result with the matching tool_use_id
  2. Dispatch each returned tool_use by its name
  3. Stop the loop on end_turn
  4. Check tool arguments against a schema

Explanation

Correct Answer

B:

The harness validates the tool identifier and arguments before invoking the corresponding local function; the important constraint is that the content block is a request to run a tool, not the executed inventory result.

Incorrect Answers

A:

The caller adds a user-side tool_result block tied to the earlier tool use identifier, whereas the requirement calls for another behavior: the harness validates the tool identifier and arguments before invoking the corresponding local function; an unrelated message cannot reliably bind a result to the invocation it completes.

C:

The harness treats the completion reason as the signal to deliver the final answer and exit; the requirement instead turns on this behavior: the harness validates the tool identifier and arguments before invoking the corresponding local function; repeating requests after natural completion can create unnecessary tool cycles.

D:

Typed validation rejects malformed or out-of-policy arguments before backend execution; this cannot substitute for the behavior that the harness validates the tool identifier and arguments before invoking the corresponding local function, because a model-generated JSON object can be syntactically valid while still violating business rules.

Question 254

A difficult architecture recommendation combines conflicting requirements, a long context, and nuanced tradeoffs. Which option best fits the described workflow?

  1. Use a capable tier for complex synthesis
  2. Use a fast tier for request classification
  3. Use a specialized deterministic tool instead of any model
  4. Run a canary rollout for a new model version

Explanation

Correct Answer

A:

A stronger model may justify its cost when deeper synthesis materially improves decisions; the implementation must still recognize that its result still requires evidence and domain-specific review.

Incorrect Answers

B:

A small capable classifier avoids spending top-tier reasoning cost on simple routing; the missing capability is more specific: a stronger model may justify its cost when deeper synthesis materially improves decisions; poor routing accuracy can still misdirect expensive follow-up steps.

C:

Trusted software calculates deterministic numeric results independently of language-model tier; by comparison, the intended solution relies on this: a stronger model may justify its cost when deeper synthesis materially improves decisions; upgrading to a larger model does not supply an arithmetic correctness guarantee.

D:

Limited exposure allows observation of quality, refusal, latency, and tool usage before full migration; the needed capability is different—a stronger model may justify its cost when deeper synthesis materially improves decisions—and a model upgrade can affect the entire user experience even when code is unchanged.

Question 255

A proposed integration may process regulated attachments but must follow enterprise storage and deletion rules. Which solution should the engineer adopt?

  1. Document data-handling and retention constraints
  2. Write testable functional acceptance criteria
  3. Identify external integration contracts
  4. Capture explicit nonfunctional latency targets

Explanation

Correct Answer

A:

Privacy requirements identify permitted collection, transmission, storage, retention, and deletion controls, although developer convenience cannot override legal or organizational handling obligations.

Incorrect Answers

B:

Requirements should specify observable inputs, citation fields, and acceptance tests for the behavior; however, the critical mechanism here is that privacy requirements identify permitted collection, transmission, storage, retention, and deletion controls, and a statement that the system should be helpful is not independently verifiable.

C:

Interface requirements document endpoints, authentication, payloads, limits, and failure handling; the requirement instead turns on this behavior: privacy requirements identify permitted collection, transmission, storage, retention, and deletion controls; a prose-level prompt is not a substitute for a validated service contract.

D:

A latency requirement defines measurable service-level timing and informs streaming and model selection; this cannot substitute for the behavior that privacy requirements identify permitted collection, transmission, storage, retention, and deletion controls, because correct semantic output does not prove the product meets a response-time constraint.

Question 256

A remote service repeatedly fails and queued work begins consuming every available thread. What should the team do at this decision point?

  1. Capture sanitized HTTP diagnostics
  2. Use a circuit breaker for an unstable dependency
  3. Differentiate SDK exceptions from business validation errors
  4. Use request timeouts for bounded waits

Explanation

Correct Answer

B:

Temporarily opening the breaker limits cascading failures and permits controlled recovery testing; a crucial qualification is that it cannot fix a bug in the caller's own payload format.

Incorrect Answers

A:

Structured redacted logs reveal failure context without leaking authorization headers; here, the controlling capability is that temporarily opening the breaker limits cascading failures and permits controlled recovery testing, but recording entire raw requests can expose customer secrets.

C:

Error categories separate transport faults from application decision failures; by comparison, the intended solution relies on this: temporarily opening the breaker limits cascading failures and permits controlled recovery testing; retry middleware should not hide an authorization denial as a transient API issue.

D:

Client networking deadlines convert unbounded waits into controlled failures; the needed capability is different—temporarily opening the breaker limits cascading failures and permits controlled recovery testing—and a model token cap does not constrain socket wait times.

Question 257

A production release must not accidentally inherit a developer's permissive local tool settings. What would best address the stated failure mode?

  1. Store environment-specific endpoints in configuration
  2. Run a configuration-diff preflight
  3. Link commits to deployed build artifacts
  4. Retain a tested rollback version

Explanation

Correct Answer

B:

Automated checks compare effective values against the approved deployment baseline, subject to this requirement: assuming staging and production are identical is not a verification method.

Incorrect Answers

A:

Deployment configuration injects environment values without rewriting application logic; in contrast, automated checks compare effective values against the approved deployment baseline, while hardcoded production URLs increase the chance of accidental test writes.

C:

Immutable artifact identifiers connect deployment state to reviewed source revisions; here, the controlling capability is that automated checks compare effective values against the approved deployment baseline, but a human-readable release label alone may not uniquely identify deployed bytes.

D:

Keeping the previously tested artifact enables a measured rollback when safe; the missing capability is more specific: automated checks compare effective values against the approved deployment baseline; a rollback should consider state migrations and outstanding jobs.

Question 258

A support agent needs to update a ticket status through an auditable backend. Which capability is necessary for this outcome?

  1. Use a sandbox for model-requested code execution
  2. Use structured tool schemas for business actions
  3. Use a custom client tool for a private business operation
  4. Limit external tools to necessary operations

Explanation

Correct Answer

B:

Typed schemas and dispatcher checks convert an intention into a constrained operation; the implementation must still recognize that general Skill text cannot replace a server-enforced action interface.

Incorrect Answers

A:

Process and network isolation contain potential side effects of generated programs; this addresses a separate concern, not the fact that typed schemas and dispatcher checks convert an intention into a constrained operation; a model's internal safety judgment is not an OS permission boundary.

C:

A user-defined tool schema plus local dispatcher safely bridges the model to that system; the missing capability is more specific: typed schemas and dispatcher checks convert an intention into a constrained operation; the developer must operate the function and enforce its permissions.

D:

Minimal capabilities reduce the harm of prompt injection or mistaken planning; by comparison, the intended solution relies on this: typed schemas and dispatcher checks convert an intention into a constrained operation; more available tools do not guarantee better task performance.

Question 259

A team wants to know whether more reasoning helps an extraction workflow. What is the best fit for the described situation?

  1. Use a lighter reasoning configuration for trivial tasks
  2. Use extra reasoning on high-complexity tasks
  3. Benchmark thinking settings with task-specific tests
  4. Use low-temperature settings when supported for consistency

Explanation

Correct Answer

C:

Controlled quality and latency measurements show whether the chosen task benefits—with this limitation: assuming maximal thinking is universally optimal wastes resources.

Incorrect Answers

A:

Lower effort on supported models is appropriate for a narrowly specified routine transformation, whereas the requirement calls for another behavior: controlled quality and latency measurements show whether the chosen task benefits; under-allocation can hurt difficult exceptions, so evaluate the change.

B:

Supported thinking or effort settings allocate greater reasoning capacity to difficult deliberation; however, the critical mechanism here is that controlled quality and latency measurements show whether the chosen task benefits, and added reasoning can increase latency and cost and still requires factual validation.

D:

Lower randomness can reduce sampling variability in supported configurations; this cannot substitute for the behavior that controlled quality and latency measurements show whether the chosen task benefits, because it does not guarantee deterministic output or eliminate model errors.

Question 260

A repository assistant is limited to read-only operations and should never see a deployment-write capability. Which response is most defensible under these conditions?

  1. Use tool allowlists as a baseline defense
  2. Use a mandatory human approval mechanism
  3. Use a pre-tool hook for prohibited actions
  4. Use constrained tool schemas and allowlists

Explanation

Correct Answer

A:

Tool availability and credential permissions restrict the reachable attack surface by design; in production, hooks are additional safeguards, not replacements for minimizing available power.

Incorrect Answers

B:

An enforced approval boundary stops execution until confirmation is provided; the needed capability is different—tool availability and credential permissions restrict the reachable attack surface by design—and a model's promise to ask first is not itself a technical approval gate.

C:

A pre-execution hook can inspect and block the operation before it reaches the shell, whereas the requirement calls for another behavior: tool availability and credential permissions restrict the reachable attack surface by design; the hook runs outside the model's judgment, but its own policy must be correctly implemented.

D:

Narrow tool capabilities reduce what even a manipulated model could attempt; however, the critical mechanism here is that tool availability and credential permissions restrict the reachable attack surface by design, and a policy paragraph in system instructions is not a technical allowlist.

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