Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 25

Question 240

A customer assistant needs to look up an order through a controlled function rather than describe a made-up query. What is the correct engineering decision?

  1. Use an explicit tool definition for callable business actions
  2. Use a user content block for the actual request
  3. Use CLAUDE.md for repository guidance in Claude Code
  4. Use a dedicated tool adapter for private systems

Explanation

Correct Answer

A:

Tool schemas advertise an available operation and define its structured inputs; a prose instruction mentioning a function does not register it with the API.

Incorrect Answers

B:

A user message carries the current request as task data subordinate to the system rules, whereas the requirement calls for another behavior: tool schemas advertise an available operation and define its structured inputs; treating customer input as a new system instruction would invert the intended trust hierarchy.

C:

Repository instruction files provide persistent project context to the Claude Code environment; however, the critical mechanism here is that tool schemas advertise an available operation and define its structured inputs, and they are not automatically loaded into unrelated Messages API clients.

D:

An adapter exposes narrowly typed operations and translates requests into the legacy interface; the requirement instead turns on this behavior: tool schemas advertise an available operation and define its structured inputs; the model does not acquire direct private-network access by seeing the system's name.

Question 241

A generated response is converted into an operational command affecting customer records. Which approach targets the deciding condition?

  1. Use constrained tool schemas and allowlists
  2. Inspect outputs before taking business actions
  3. Enforce least-privilege infrastructure permissions
  4. Keep a human review gate for high-impact operations

Explanation

Correct Answer

B:

Output checks validate schema, scope, and safety before any backend mutation; the important constraint is that post-generation checks cannot repair unauthorized access already performed upstream.

Incorrect Answers

A:

Narrow tool capabilities reduce what even a manipulated model could attempt, whereas the requirement calls for another behavior: output checks validate schema, scope, and safety before any backend mutation; a policy paragraph in system instructions is not a technical allowlist.

C:

Backend credentials and IAM policies restrict it to the minimum necessary operations; the requirement instead turns on this behavior: output checks validate schema, scope, and safety before any backend mutation; an unrestricted credential remains dangerous regardless of the model's stated intent.

D:

Mandatory human approval separates recommendations from irreversible execution; this cannot substitute for the behavior that output checks validate schema, scope, and safety before any backend mutation, because model confidence scores cannot replace the required authority of a reviewer.

Question 242

A web app needs a one-way stream of generated assistant text over an HTTP connection. Which solution matches the actual constraint?

  1. Use request timeouts for bounded waits
  2. Use server-sent events for streamed responses
  3. Use a queue for disconnected long-running workloads
  4. Use HTTP request-response for discrete inference

Explanation

Correct Answer

B:

SSE carries incremental events from the service to the client as generation proceeds, although it is not a general-purpose bidirectional client messaging transport.

Incorrect Answers

A:

Client networking deadlines convert unbounded waits into controlled failures; the needed capability is different—SSE carries incremental events from the service to the client as generation proceeds—and a model token cap does not constrain socket wait times.

C:

Durable queued work survives client disconnection and exposes later status; however, the critical mechanism here is that SSE carries incremental events from the service to the client as generation proceeds, and keeping an open WebSocket alone does not make background jobs persistent.

D:

Standard HTTPS requests fit isolated generation operations without long-lived bidirectional connections; the requirement instead turns on this behavior: SSE carries incremental events from the service to the client as generation proceeds; opening a persistent socket adds state and lifecycle complexity where it is unnecessary.

Question 243

A tool call is rejected because the agent lacks access to a protected customer account. Which of these alternatives serves the stated goal?

  1. Retry a failed read with bounded backoff
  2. Use a compensating action for a committed write
  3. Finish with a user-facing answer after the last action
  4. Abandon an unauthorized operation immediately

Explanation

Correct Answer

D:

Denial should terminate that action or route it through legitimate approval instead of repeat attempts; retrying with longer waits cannot transform insufficient permission into authorization.

Incorrect Answers

A:

Retrying after increasing delays can recover a transient read failure without changing external state; here, the controlling capability is that denial should terminate that action or route it through legitimate approval instead of repeat attempts, but the caller must stop after a limited number of attempts and surface persistent failure.

B:

A compensating operation undoes a completed business change when atomic rollback spans unavailable services; the missing capability is more specific: denial should terminate that action or route it through legitimate approval instead of repeat attempts; compensation must itself be auditable and may require authorized human review.

C:

The model synthesizes a final response once no additional client tool request is needed; by comparison, the intended solution relies on this: denial should terminate that action or route it through legitimate approval instead of repeat attempts; a raw tool result is not automatically a contextualized customer answer.

Question 244

Claude asks to query an internal database through a user-defined tool. What is the most suitable pattern?

  1. Correlate batch results by custom identifiers
  2. Handle multiple tool_use blocks as separate requests
  3. Register a client tool with input_schema
  4. Execute the returned tool_use in application code

Explanation

Correct Answer

D:

The integrator validates and runs the operation locally, then returns the result, although the Claude API cannot directly call arbitrary internal customer functions by itself.

Incorrect Answers

A:

Stable request IDs let the application map each result to the correct original record; the missing capability is more specific: the integrator validates and runs the operation locally, then returns the result; relying on result ordering can mix customer records or output destinations.

B:

The dispatcher processes and maps every permitted call by its own identifier before continuation; by comparison, the intended solution relies on this: the integrator validates and runs the operation locally, then returns the result; returning a result for only one requested call leaves unresolved tool work.

C:

Tool definitions describe names, parameters, and types so Claude can propose structured calls; the needed capability is different—the integrator validates and runs the operation locally, then returns the result—and defining a schema does not grant access to an external system automatically.

Question 245

An organization operates both staging and production Claude integrations. It must avoid committing production keys, detect unusual credential use, and keep nonproduction jobs away from live quota. Which THREE credential controls should it combine? Choose THREE.

  1. Store credentials in a secret manager
  2. Monitor key usage anomalies
  3. Use separate credentials for development and production
  4. Require consent for delegated third-party scopes
  5. Audit approval decisions for high-impact actions
  6. Rotate a key after suspected leakage

Explanation

Correct Answers

A:

Protected storage and runtime injection keep billable API keys out of committed files; its operational boundary is that encryption at rest does not justify exposing secrets in source control.

B:

Usage attribution, budget alerts, and access logs can expose unauthorized patterns for investigation, subject to this requirement: a key that still works is not evidence that its use is legitimate.

C:

Environment isolation makes accidental or compromised test execution less damaging; the implementation must still recognize that a single unrestricted key creates unnecessary blast radius across environments.

Incorrect Answers

D:

A proper authorization and consent flow limits the scopes granted to the integration; the missing capability is more specific: protected storage and runtime injection keep billable API keys out of committed files; model-generated text cannot silently supply the user's valid consent.

E:

A durable record captures the approver, scope, timestamp, and action approved; by comparison, the intended solution relies on this: protected storage and runtime injection keep billable API keys out of committed files; an assistant's paraphrase of approval is not equivalent to an auditable authorization event.

F:

Revoking and replacing the affected secret reduces the time attackers can continue using it; the needed capability is different—protected storage and runtime injection keep billable API keys out of committed files—and hiding the log page alone cannot guarantee the secret was not captured.

Question 246

A single customer query triggers multiple inference and tool calls across microservices. Which of the following would work as specified?

  1. Correlate downstream calls with request IDs
  2. Redact credentials before logging
  3. Escalate repeated unknown errors with trace context
  4. Record latency per integration stage

Explanation

Correct Answer

A:

Propagated trace identifiers let operators reconstruct a request's causal path; its operational boundary is that comparing timestamps manually is unreliable under concurrency.

Incorrect Answers

B:

Logging middleware removes secrets before storage or export to observability systems; this addresses a separate concern, not the fact that propagated trace identifiers let operators reconstruct a request's causal path; restricting dashboard viewers does not erase sensitive fields already written.

C:

Capturing sanitized request IDs and relevant version metadata makes specialist diagnosis possible; here, the controlling capability is that propagated trace identifiers let operators reconstruct a request's causal path, but silently dropping the incident gives no route to correction.

D:

Timed spans separate request phases and identify the bottleneck for targeted optimization; the missing capability is more specific: propagated trace identifiers let operators reconstruct a request's causal path; total elapsed time alone cannot localize the slow dependency.

Question 247

A requirements analyst must finish constraints before an implementer can draft a technical design. Which change would satisfy the constraint?

  1. Split by sequential expertise
  2. Avoid splitting when coordination overhead dominates
  3. Split by security boundary
  4. Use typed PydanticAI-style dependencies

Explanation

Correct Answer

A:

A staged handoff gives each specialist the prior approved artifact it needs; parallelizing dependent expert roles would force the implementer to guess requirements.

Incorrect Answers

B:

Keeping a bounded task local avoids orchestration latency and reconciliation errors, whereas the requirement calls for another behavior: a staged handoff gives each specialist the prior approved artifact it needs; subagents should be justified by distinct work or isolation needs, not mere availability.

C:

Isolation by trust zone prevents a public-facing worker from gaining private data access; however, the critical mechanism here is that a staged handoff gives each specialist the prior approved artifact it needs, and worker separation without separate credentials or tool allowlists is superficial.

D:

A typed framework aligns model interactions with application data models and testable dependencies; the requirement instead turns on this behavior: a staged handoff gives each specialist the prior approved artifact it needs; types validate shape, not the truth of every generated factual assertion.

Question 248

A test pipeline should not be able to spend production quota or access live customer integrations. What should the team do at this decision point?

  1. Store credentials in a secret manager
  2. Use separate credentials for development and production
  3. Monitor key usage anomalies
  4. Audit approval decisions for high-impact actions

Explanation

Correct Answer

B:

Environment isolation makes accidental or compromised test execution less damaging; notably, a single unrestricted key creates unnecessary blast radius across environments.

Incorrect Answers

A:

Protected storage and runtime injection keep billable API keys out of committed files; the requirement instead turns on this behavior: environment isolation makes accidental or compromised test execution less damaging; encryption at rest does not justify exposing secrets in source control.

C:

Usage attribution, budget alerts, and access logs can expose unauthorized patterns for investigation; in contrast, environment isolation makes accidental or compromised test execution less damaging, while a key that still works is not evidence that its use is legitimate.

D:

A durable record captures the approver, scope, timestamp, and action approved; this addresses a separate concern, not the fact that environment isolation makes accidental or compromised test execution less damaging; an assistant's paraphrase of approval is not equivalent to an auditable authorization event.

Question 249

A client needs to browse a maintained policy document without treating the document itself as a callable operation. What is the right implementation response?

  1. Expose an MCP resource for read-oriented context
  2. Use stdio for a local subprocess integration
  3. Expose an MCP prompt for a reusable template
  4. Expose an MCP tool for an executable operation

Explanation

Correct Answer

A:

Resources represent accessible data that a client can retrieve and supply as context; a crucial qualification is that a resource's text cannot authorize additional tool actions by itself.

Incorrect Answers

B:

Standard input and output suit tightly coupled local client-server interactions; by comparison, the intended solution relies on this: resources represent accessible data that a client can retrieve and supply as context; the server's standard output must remain protocol-clean instead of mixing human debug prints.

C:

Prompt templates provide reusable structured instructions rather than executing business operations; the needed capability is different—resources represent accessible data that a client can retrieve and supply as context—and templates do not automatically grant access to any tool or private resource.

D:

An MCP tool advertises callable functionality with input schemas and returns execution results, whereas the requirement calls for another behavior: resources represent accessible data that a client can retrieve and supply as context; tool calls still need explicit runtime permissions and backend validation.

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.

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