Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 42

Question 410

An assistant reads a web page that includes instructions telling it to send secret credentials elsewhere. Which response is most defensible under these conditions?

  1. Monitor retrieval coverage and citation fidelity
  2. Keep user instructions subordinate to security policy
  3. Use tool results as data to be interpreted
  4. Treat retrieved pages as untrusted evidence

Explanation

Correct Answer

D:

External page text should be handled as data, not as new operational authority; the important constraint is that a page author cannot legitimately override the application's system policy.

Incorrect Answers

A:

Measuring evidence relevance and attribution identifies grounding defects; by comparison, the intended solution relies on this: external page text should be handled as data, not as new operational authority; extra context tokens cannot guarantee faithful citation of the retrieved material.

B:

The application enforces identity and authorization checks despite the request's wording; the needed capability is different—external page text should be handled as data, not as new operational authority—and requests from end users do not grant new backend privileges.

C:

Result content is evidence from an external subsystem and must not become a system instruction, whereas the requirement calls for another behavior: external page text should be handled as data, not as new operational authority; even trusted services can carry attacker-controlled record values.

Question 411

A completed custom tool operation needs to supply its data back for the assistant's next step. Which of the following would work as specified?

  1. Read the returned tool_use ID and name
  2. Use an Anthropic server-executed tool where supported
  3. Validate and execute the tool outside the model
  4. Return a matching tool_result block

Explanation

Correct Answer

D:

The following turn includes a result linked by tool_use_id so the model can continue coherently; a crucial qualification is that returning an unrelated human message loses the execution association.

Incorrect Answers

A:

Each client tool request includes metadata that ties subsequent execution and results to that call; in contrast, the following turn includes a result linked by tool_use_id so the model can continue coherently, while guessing an operation from nearby text can dispatch the wrong function.

B:

Server tools execute on the service side and return associated results during the model interaction; this addresses a separate concern, not the fact that the following turn includes a result linked by tool_use_id so the model can continue coherently; they have distinct data and capability boundaries from the customer's custom functions.

C:

Application code enforces permission checks and performs the actual side effect or read; here, the controlling capability is that the following turn includes a result linked by tool_use_id so the model can continue coherently, but the presence of tool_use content is not evidence the backend was queried.

Question 412

A multi-document prompt exceeds the selected model's supported context capacity. What is the most suitable pattern?

  1. Recognize latency due to extra reasoning
  2. Handle context overflow before submission
  3. Count the complete request against context limits
  4. Bound total cost across agent turns

Explanation

Correct Answer

B:

Chunking, retrieval, or compaction reduces input size while retaining decision-relevant evidence—with this limitation: increasing output max_tokens cannot make an overlong input fit.

Incorrect Answers

A:

Reasoning controls and task complexity affect total latency independently of answer length; however, the critical mechanism here is that chunking, retrieval, or compaction reduces input size while retaining decision-relevant evidence, and reducing displayed text alone may not remove the underlying thinking cost.

C:

The context budget encompasses the relevant input contents and model-specific output allowances; this cannot substitute for the behavior that chunking, retrieval, or compaction reduces input size while retaining decision-relevant evidence, because counting only the latest user text understates true token pressure.

D:

Aggregating tokens and actions by logical task reveals overall expenditure and quota needs; in contrast, chunking, retrieval, or compaction reduces input size while retaining decision-relevant evidence, while per-call estimates overlook the cumulative work of multi-turn orchestration.

Question 413

A proposed customer refund should be calculated by the app but posted only after authorized consent. Which mechanism should the team prioritize?

  1. Use a custom client tool for a private business operation
  2. Combine a deterministic hook with a coding agent
  3. Combine a guarded custom tool with a human approval step
  4. Combine a Skill with a read-only tool

Explanation

Correct Answer

C:

The tool executes a constrained operation after an enforced business approval gate; model-generated consent text is not an auditable approval event.

Incorrect Answers

A:

A user-defined tool schema plus local dispatcher safely bridges the model to that system; the missing capability is more specific: the tool executes a constrained operation after an enforced business approval gate; the developer must operate the function and enforce its permissions.

B:

A hook enforces the external check while the model handles the flexible development task; by comparison, the intended solution relies on this: the tool executes a constrained operation after an enforced business approval gate; without correct hook wiring and tool permissions, guidance alone is insufficient.

D:

The Skill supplies the workflow while the tool supplies authorized external facts, whereas the requirement calls for another behavior: the tool executes a constrained operation after an enforced business approval gate; instruction packages and executable integrations solve different needs.

Question 414

Users can send giant attachments that consume memory before reaching the model. Which change would satisfy the constraint?

  1. Apply backpressure to input queues
  2. Reject unexpected fields in business actions
  3. Enforce upload-size caps
  4. Set client-side network timeouts

Explanation

Correct Answer

C:

Boundary checks reject overly large payloads before expensive parsing or API transfer; a crucial qualification is that relying on the model to refuse oversized data is too late for resource protection.

Incorrect Answers

A:

Backpressure or queue admission control protects worker memory and service availability; this addresses a separate concern, not the fact that boundary checks reject overly large payloads before expensive parsing or API transfer; unlimited buffering merely postpones overload until the system fails.

B:

Application validation controls the allowed property set before any database mutation; here, the controlling capability is that boundary checks reject overly large payloads before expensive parsing or API transfer, but the model may produce syntactically valid extra keys that are unsafe to execute.

D:

Connect and read deadlines bound network waits and allow controlled failure behavior; by comparison, the intended solution relies on this: boundary checks reject overly large payloads before expensive parsing or API transfer; token limits describe model output, not HTTP socket lifetime.

Question 415

A model is told to keep outputs short and also include all legally required warnings. Which change would satisfy the constraint?

  1. State the output task and audience explicitly
  2. Use explicit delimiters for source passages
  3. Prioritize conflicting constraints in a policy
  4. Separate requirements from source data

Explanation

Correct Answer

C:

An explicit hierarchy identifies which requirements take precedence when they conflict; unresolved contradictory instructions can produce unpredictable omissions.

Incorrect Answers

A:

Clear instructions define the task, intended reader, and required level of technical detail; the missing capability is more specific: an explicit hierarchy identifies which requirements take precedence when they conflict; a vague request to be useful cannot anchor a reliable output format.

B:

Marking the passage's bounds clarifies which content is data rather than a user command; by comparison, the intended solution relies on this: an explicit hierarchy identifies which requirements take precedence when they conflict; delimiters help interpretation but do not replace tool-level security.

D:

Structural boundaries distinguish data to analyze from commands the assistant should follow, whereas the requirement calls for another behavior: an explicit hierarchy identifies which requirements take precedence when they conflict; external content should not gain authority merely because it appears near instructions.

Question 416

A request contains an invalid project identifier and the client needs to correct it. Which mechanism should the team prioritize?

  1. Return actionable validation errors to callers
  2. Escalate repeated unknown errors with trace context
  3. Count errors by class and endpoint
  4. Use a safe degraded response on dependency outage

Explanation

Correct Answer

A:

A structured 4xx response names the offending field without leaking internal implementation details; a generic 500 would misleadingly suggest a server outage.

Incorrect Answers

B:

Capturing sanitized request IDs and relevant version metadata makes specialist diagnosis possible, whereas the requirement calls for another behavior: a structured 4xx response names the offending field without leaking internal implementation details; silently dropping the incident gives no route to correction.

C:

Structured metrics reveal whether the issue is client validation, quota pressure, or upstream availability; however, the critical mechanism here is that a structured 4xx response names the offending field without leaking internal implementation details, and a single error counter masks very different remediation paths.

D:

A fallback can queue the request or return a limited verified response without fabricating model results; the requirement instead turns on this behavior: a structured 4xx response names the offending field without leaking internal implementation details; retrying indefinitely inside the user's HTTP connection harms availability.

Question 417

Independent research workers must return findings with the same fields for a reliable consolidated report. Which mechanism should the team prioritize?

  1. Use a reducer with a fixed schema
  2. Propagate unresolved uncertainties
  3. Deduplicate overlapping citations and actions
  4. Retry a failed read with bounded backoff

Explanation

Correct Answer

A:

A merger validates and combines structured outputs rather than concatenating arbitrary prose; notably, different worker vocabularies can break aggregation unless normalized first.

Incorrect Answers

B:

The aggregate output carries gaps and confidence limitations from each contributor; in contrast, a merger validates and combines structured outputs rather than concatenating arbitrary prose, while a polished synthesis should not erase unsupported assumptions.

C:

Deduplication by stable identity avoids repeated evidence and duplicate business actions; this addresses a separate concern, not the fact that a merger validates and combines structured outputs rather than concatenating arbitrary prose; text similarity alone is not enough to safely deduplicate side effects.

D:

Retrying after increasing delays can recover a transient read failure without changing external state; here, the controlling capability is that a merger validates and combines structured outputs rather than concatenating arbitrary prose, but the caller must stop after a limited number of attempts and surface persistent failure.

Question 418

A CI coding-agent job only needs to inspect diffs but its default environment contains deployment tools. Which production safeguard is most relevant?

  1. Use a custom slash command for a named recurring action
  2. Limit shell command authority in automation
  3. Review effective settings before broad automation
  4. Resume a relevant prior session intentionally

Explanation

Correct Answer

B:

Restricted command/tool permissions reduce the possible effects of prompt errors or malicious code comments; the implementation must still recognize that using a temporary checkout is helpful but does not itself revoke external credentials.

Incorrect Answers

A:

A named command exposes reusable prompt-driven workflow entry points inside Claude Code; this addresses a separate concern, not the fact that restricted command/tool permissions reduce the possible effects of prompt errors or malicious code comments; a command definition still depends on the session's tool permissions and context.

C:

Examining effective permission and hook configuration catches dangerous inherited allowances; the missing capability is more specific: restricted command/tool permissions reduce the possible effects of prompt errors or malicious code comments; a printed prompt cannot override an enforced policy hierarchy.

D:

Session resumption restores applicable prior conversational context for continuation; by comparison, the intended solution relies on this: restricted command/tool permissions reduce the possible effects of prompt errors or malicious code comments; resuming the wrong session can import stale assumptions or sensitive unrelated history.

Question 419

A model suggests a query for a database holding financial records and the app must avoid arbitrary execution. Which option best fits the described workflow?

  1. Require approval for write statements
  2. Parse generated JSON with a real parser
  3. Execute generated code only in a restricted sandbox
  4. Treat generated SQL as untrusted text

Explanation

Correct Answer

D:

Query validation and read-only permissions constrain execution against approved tables and patterns; a crucial qualification is that text generation does not certify the safety or accuracy of the SQL.

Incorrect Answers

A:

A controlled executor enforces authorized write permissions and human approval where required; in contrast, query validation and read-only permissions constrain execution against approved tables and patterns, while a syntax-valid UPDATE statement is not proof of business permission.

B:

Standard JSON parsing catches syntactic errors instead of relying on visual plausibility; this addresses a separate concern, not the fact that query validation and read-only permissions constrain execution against approved tables and patterns; successfully parsed JSON can still violate expected business types.

C:

Sandboxing limits filesystem, network, process, and credential exposure; here, the controlling capability is that query validation and read-only permissions constrain execution against approved tables and patterns, but asking the model to promise safety is not an operating-system isolation mechanism.

Question 420

An integration is migrating between Claude model generations that expose different reasoning controls. What is the appropriate next design choice?

  1. Choose a supported model identifier
  2. Check model compatibility before changing thinking mode
  3. Budget total latency for thinking and output
  4. Use supported thinking controls for a reasoning task

Explanation

Correct Answer

B:

A capability check and test run prevent invalid parameter combinations and behavioral regressions; notably, assuming every family supports the same thinking configuration is unsafe.

Incorrect Answers

A:

The model field identifies the deployed model; production should verify current IDs and compatibility; the requirement instead turns on this behavior: a capability check and test run prevent invalid parameter combinations and behavioral regressions; guessing an informal nickname can yield an invalid-model error or unexpected alias changes.

C:

Performance tests include model reasoning time as well as visible answer generation; in contrast, a capability check and test run prevent invalid parameter combinations and behavioral regressions, while restricting just the final response length may not address total computation time.

D:

Request settings trade token consumption and latency against potential reasoning performance; this addresses a separate concern, not the fact that a capability check and test run prevent invalid parameter combinations and behavioral regressions; thinking support and parameter syntax depend on the particular model and API revision.

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