Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 32

Question 310

A team maintaining a sensitive-data assistant must meet 2 separate acceptance criteria: (1) detect unauthorized modifications to a tool result before it reaches the model; (2) stop frontend users from receiving unrestricted backend tool credentials. Which two implementation actions directly satisfy the stated requirements? Choose TWO.

  1. Pause execution for explicit approval at the high-impact action boundary.
  2. Verify origin and integrity of sensitive external data where supported.
  3. Use a server-side workload identity or scoped service credential.
  4. Treat retrieved content as untrusted and prevent its instructions from controlling tools.
  5. Isolate external search results and validate any resulting tool action independently.

Explanation

Correct Answers

B:

The test to detect unauthorized modifications to a tool result before it reaches the model is met because trusted decisions require confidence that tool outputs were not altered; transport and signature checks complement content validation; the remaining selection covers the separate need to stop frontend users from receiving unrestricted backend tool credentials.

C:

The test to stop frontend users from receiving unrestricted backend tool credentials is met because service identities tie privileges to a runtime workload and avoid exposing long-lived secrets to browsers or untrusted users; the remaining selection covers the separate need to detect unauthorized modifications to a tool result before it reaches the model.

Incorrect Answers

A:

The action is useful for the separate capability to require a person to approve an irreversible administrative change because human approval is a policy-enforced transition, not a note written in the model's plan or a message displayed afterward; however, the required test outcomes are to detect unauthorized modifications to a tool result before it reaches the model and to stop frontend users from receiving unrestricted backend tool credentials.

D:

The action is useful for the separate capability to resist instructions embedded in a retrieved support ticket that demand secret disclosure because source material is data rather than authority; enforcing tool permissions and separating roles limits an attacker's ability to redirect execution; however, the required test outcomes are to detect unauthorized modifications to a tool result before it reaches the model and to stop frontend users from receiving unrestricted backend tool credentials.

E:

The action is useful for the separate capability to reject a malicious instruction inserted into a web-search result rather than the user's direct request because indirect attacks exploit lower-trust retrieval surfaces, so trusting search snippets as operational directives crosses a security boundary; however, the required test outcomes are to detect unauthorized modifications to a tool result before it reaches the model and to stop frontend users from receiving unrestricted backend tool credentials.

Question 311

A team maintaining an autonomous repository maintenance job has two different engineering requirements: (1) coordinate several specialized investigations and assemble one accountable final answer, and (2) stop iterative tool use when a success condition or resource budget is reached. Which TWO independent changes directly meet them? Choose TWO.

  1. Use isolated branch state and a controlled reconciliation step.
  2. Encode a deterministic workflow with explicit transitions between stages.
  3. Use a supervisor agent to delegate bounded tasks and synthesize results.
  4. Enforce explicit completion checks and a maximum iteration budget.
  5. Run independent branches concurrently and merge their outputs afterward.

Explanation

Correct Answers

C:

To coordinate several specialized investigations and assemble one accountable final answer as part of a design that must also stop iterative tool use when a success condition or resource budget is reached, a supervisor owns the overall objective and integrates subagent findings; each delegate should have a narrow mandate and return a constrained summary; this is why the action is one of the two required selections.

D:

To stop iterative tool use when a success condition or resource budget is reached as part of a design that must also coordinate several specialized investigations and assemble one accountable final answer, bounded loops avoid unending tool calls and uncontrolled spend while making incomplete work visible instead of silently continuing; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to stop two parallel branches from overwriting the same mutable record because explicit merge rules are required when outputs converge; concurrent mutation without arbitration causes nondeterministic results; the required outcomes are instead to coordinate several specialized investigations and assemble one accountable final answer and to stop iterative tool use when a success condition or resource budget is reached.

B:

This alternative serves the separate goal to run the same retrieval, validation, and publishing stages in an approved order for every request because known control flow belongs in application code, allowing each stage to be tested and audited without asking a model to improvise the sequence; the required outcomes are instead to coordinate several specialized investigations and assemble one accountable final answer and to stop iterative tool use when a success condition or resource budget is reached.

E:

This alternative serves the separate goal to reduce elapsed time for independent extraction tasks that do not share mutable state because parallel orchestration shortens the critical path when calls have no dependencies, while the merger checks completeness before proceeding; the required outcomes are instead to coordinate several specialized investigations and assemble one accountable final answer and to stop iterative tool use when a success condition or resource budget is reached.

Question 312

In a Claude inference cost review, a release requirement is to explain why language-model output is generated step by step rather than fetched from a fixed response table. Which implementation addresses that specific requirement?

  1. Budget total context for instructions, history, tools, and output.
  2. Use more exploratory sampling where supported and measure quality.
  3. Check each response block's type before extracting its fields.
  4. Model the response as conditioned token generation.

Explanation

Correct Answer

D:

To explain why language-model output is generated step by step rather than fetched from a fixed response table in this Claude inference cost review, autoregressive decoding selects subsequent tokens based on supplied context and prior output, which creates flexible but non-guaranteed wording; the proposed implementation directly supplies that capability.

Incorrect Answers

A:

The proposed alternative helps to know why a long prompt plus a proposed response may exceed the model's accepted context because a context window bounds what can be processed in one request; large tool results consume it along with user messages; that mechanism does not meet the different requirement to explain why language-model output is generated step by step rather than fetched from a fixed response table.

B:

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 explain why language-model output is generated step by step rather than fetched from a fixed response table.

C:

The proposed alternative helps to prevent a tool_use block from being displayed as if it were natural-language text because typed response objects can contain several block families; assuming every block has a text field causes logic errors; that mechanism does not meet the different requirement to explain why language-model output is generated step by step rather than fetched from a fixed response table.

Question 313

Two approval criteria are still open for a business workflow that calls Claude: the team must reject tool arguments that fail a declared shape before tool execution and stop a well-formed JSON response from introducing an impossible customer status. Which two implementation actions satisfy them? Choose TWO.

  1. Add application-owned approval logic around the refund tool.
  2. Give mutating operations an application-level idempotency key.
  3. Use strict tool input validation for the exposed tool interface.
  4. Place stable application rules in system instructions and untrusted requests in user content.
  5. Validate generated fields against domain rules after parsing.

Explanation

Correct Answers

C:

To reject tool arguments that fail a declared shape before tool execution as part of a design that must also stop a well-formed JSON response from introducing an impossible customer status, a strict schema constrains the arguments a model may emit, but application authorization and business-rule checks remain necessary; this is why the action is one of the two required selections.

E:

To stop a well-formed JSON response from introducing an impossible customer status as part of a design that must also reject tool arguments that fail a declared shape before tool execution, a syntactically valid object may still contain invalid business values; server-side semantics determine whether it is acceptable; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to ensure refunds above a policy limit receive human sign-off because decisions that create financial side effects require deterministic authorization checks beyond a model's recommendation; the required outcomes are instead to reject tool arguments that fail a declared shape before tool execution and to stop a well-formed JSON response from introducing an impossible customer status.

B:

This alternative serves the separate goal to prevent duplicate purchases or tickets if an agent repeats a tool call because idempotency ensures replays of the same intended action do not create duplicate real-world effects even when orchestration retries; the required outcomes are instead to reject tool arguments that fail a declared shape before tool execution and to stop a well-formed JSON response from introducing an impossible customer status.

D:

This alternative serves the separate goal to keep application policy separate from a user's question because instruction placement preserves the distinction between developer intent and user-supplied data; concatenating everything blurs that boundary; the required outcomes are instead to reject tool arguments that fail a declared shape before tool execution and to stop a well-formed JSON response from introducing an impossible customer status.

Question 314

A team maintaining a production prompt pipeline has two different engineering requirements: (1) measure whether a changed instruction actually improves extraction accuracy, and (2) maintain label semantics when a few-shot prompt contains dialogue. Which TWO independent changes directly meet them? Choose TWO.

  1. Instruct the application to request clarification for unresolved high-impact ambiguity.
  2. Ground consequential statements in retrieved evidence or external checks.
  3. Use properly labeled example inputs and outputs instead of mixing speakers.
  4. Apply domain-specific authorization and business rules after schema validation.
  5. Hold evaluation inputs constant and compare prompt versions.

Explanation

Correct Answers

C:

To maintain label semantics when a few-shot prompt contains dialogue as part of a design that must also measure whether a changed instruction actually improves extraction accuracy, correct role boundaries keep demonstrations interpretable and prevent an example answer from looking like a live instruction; this is why the action is one of the two required selections.

E:

To measure whether a changed instruction actually improves extraction accuracy as part of a design that must also maintain label semantics when a few-shot prompt contains dialogue, a reproducible prompt test isolates the change being studied; anecdotal output from unrelated cases does not show improvement; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to ask for missing essential parameters rather than inventing a consequential choice because clarification reduces irreversible guesses when requirements do not uniquely determine a safe action; the required outcomes are instead to measure whether a changed instruction actually improves extraction accuracy and to maintain label semantics when a few-shot prompt contains dialogue.

B:

This alternative serves the separate goal to avoid treating fluent model wording as evidence that a claim is verified because linguistic confidence is not a calibrated truth guarantee, so factual correctness requires independent support; the required outcomes are instead to measure whether a changed instruction actually improves extraction accuracy and to maintain label semantics when a few-shot prompt contains dialogue.

D:

This alternative serves the separate goal to block a technically well-formed refund record with an unauthorized amount because type correctness is not permission to act; amounts, account ownership, and policy limits need independent checks; the required outcomes are instead to measure whether a changed instruction actually improves extraction accuracy and to maintain label semantics when a few-shot prompt contains dialogue.

Question 315

An architecture review of a tool-using incident triage agent asks how to pursue an open-ended investigation where the next action depends on intermediate evidence. Which action is the best fit?

  1. Define and validate a typed schema at the handoff.
  2. Give an agent a bounded objective and a tool-selection loop.
  3. Run independent branches concurrently and merge their outputs afterward.
  4. Consume the agent runtime's streamed event sequence.

Explanation

Correct Answer

B:

To pursue an open-ended investigation where the next action depends on intermediate evidence in this tool-using incident triage agent, agentic control lets the model choose among permitted steps as observations arrive; budgets and termination conditions still bound its autonomy; the proposed implementation directly supplies that capability.

Incorrect Answers

A:

The proposed alternative helps to pass reliable machine-readable outputs between independently developed agent stages because typed outputs make downstream contracts inspectable and reduce ambiguity caused by free-form prose passed as control data; that mechanism does not meet the different requirement to pursue an open-ended investigation where the next action depends on intermediate evidence.

C:

The proposed alternative helps to reduce elapsed time for independent extraction tasks that do not share mutable state because parallel orchestration shortens the critical path when calls have no dependencies, while the merger checks completeness before proceeding; that mechanism does not meet the different requirement to pursue an open-ended investigation where the next action depends on intermediate evidence.

D:

The proposed alternative helps to deliver intermediate agent events to the client as they occur because event-driven consumption lets the application display progress and inspect tool actions without waiting for the entire multi-step run to finish; that mechanism does not meet the different requirement to pursue an open-ended investigation where the next action depends on intermediate evidence.

Question 316

In a model-family migration plan, testing reveals that the system cannot account for different plausible answers from the same input on repeated runs and cannot reduce inconsistent labeling across nuanced examples. Select the two targeted actions that close these gaps. Choose TWO.

  1. Design for stochastic output and evaluate behavior over representative repeats.
  2. Tune sampling for lower variance while retaining output validation.
  3. Include several carefully selected and diverse demonstrations.
  4. Place caching around stable ordered content before dynamic material.
  5. Revalidate prompts, schemas, safety constraints, and tools after model migration.

Explanation

Correct Answers

A:

To account for different plausible answers from the same input on repeated runs as part of a design that must also reduce inconsistent labeling across nuanced examples, sampling and model execution can produce variation; important correctness must be checked rather than assumed identical on every call; this is why the action is one of the two required selections.

C:

To reduce inconsistent labeling across nuanced examples as part of a design that must also account for different plausible answers from the same input on repeated runs, multiple examples communicate decision boundaries and desired formatting, although irrelevant or misleading demonstrations can bias the model; this is why the action is one of the two required selections.

Incorrect Answers

B:

This alternative serves the separate goal 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; the required outcomes are instead to account for different plausible answers from the same input on repeated runs and to reduce inconsistent labeling across nuanced examples.

D:

This alternative serves the separate goal to prevent constantly changing user text from invalidating a reusable prompt prefix because cache matching operates over the stable prefix, so volatile material before the breakpoint undermines reuse; the required outcomes are instead to account for different plausible answers from the same input on repeated runs and to reduce inconsistent labeling across nuanced examples.

E:

This alternative serves the separate goal to avoid assuming a model-version change preserves tool output and writing behavior because new releases can change response distributions or supported behaviors, making a model upgrade a real software change; the required outcomes are instead to account for different plausible answers from the same input on repeated runs and to reduce inconsistent labeling across nuanced examples.

Question 317

An end-to-end test for a multi-tenant support application checks whether it can recover transient 429 or 5xx responses without creating a retry storm and reduce repeated TLS setup overhead for high-volume application traffic. Which plan should be approved to satisfy both checks?

  1. Perform both steps: (1) Retry eligible failures with capped exponential backoff and jitter; (2) reuse an appropriately configured http client or sdk connection pool.
  2. Perform both steps: (1) Set the appropriate application/json content type and serialize a valid payload; (2) reuse an appropriately configured http client or sdk connection pool.
  3. Perform both steps: (1) Retry eligible failures with capped exponential backoff and jitter; (2) authenticate incoming callbacks and validate their signatures where supported.
  4. Perform both steps: (1) Set the appropriate application/json content type and serialize a valid payload; (2) authenticate incoming callbacks and validate their signatures where supported.

Explanation

Correct Answer

A:

The first change enables the team to recover transient 429 or 5xx responses without creating a retry storm because backoff spreads retry traffic and respects transient capacity limits; blind immediate repetition worsens overload; the second addresses the separate need to reduce repeated TLS setup overhead for high-volume application traffic because connection reuse lowers transport overhead but still requires limits and health handling for long-lived clients.

Incorrect Answers

B:

The second step meets the requirement to reduce repeated TLS setup overhead for high-volume application traffic because connection reuse lowers transport overhead but still requires limits and health handling for long-lived clients; the first step would only help to have the server parse a JSON request body predictably because a JSON body requires both correct serialization and matching HTTP metadata; raw language-model text cannot replace a valid request object, leaving the separate requirement to recover transient 429 or 5xx responses without creating a retry storm unmet.

C:

The first step is appropriate for the requirement to recover transient 429 or 5xx responses without creating a retry storm because backoff spreads retry traffic and respects transient capacity limits; blind immediate repetition worsens overload; the other step instead aims 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, which does not meet the second requirement to reduce repeated TLS setup overhead for high-volume application traffic.

D:

Both steps target different results—have the server parse a JSON request body predictably and trust a callback only after verifying who actually sent it—because a JSON body requires both correct serialization and matching HTTP metadata; raw language-model text cannot replace a valid request object and network arrival alone does not prove a webhook's origin; authenticity checks prevent forged events from changing agent state; the required capabilities are instead to recover transient 429 or 5xx responses without creating a retry storm and to reduce repeated TLS setup overhead for high-volume application traffic.

Question 318

In a repository-level Claude Code rollout, testing reveals that the system cannot delegate an independent code review without filling the main coding session with details and cannot continue a previously interrupted Claude Code work session. Select the two targeted actions that close these gaps. Choose TWO.

  1. Create a project skill with the documented skill metadata and content.
  2. Configure a specialized Claude Code subagent with a narrow remit.
  3. Use the supported session continuation or resume mechanism.
  4. Add concise project guidance to the appropriate CLAUDE.md file.
  5. Define policy in the appropriate managed or shared settings scope.

Explanation

Correct Answers

B:

To delegate an independent code review without filling the main coding session with details as part of a design that must also continue a previously interrupted Claude Code work session, subagents can focus on a task and return a summary, limiting noise in the parent session and allowing separate permissions when configured; this is why the action is one of the two required selections.

C:

To continue a previously interrupted Claude Code work session as part of a design that must also delegate an independent code review without filling the main coding session with details, restoring session context reduces lost work while preserving continuity of prior tool activity; opening a new unrelated session does not do the same; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to make a reusable development procedure available to Claude Code with supporting files because a skill packages guidance or a workflow for reuse across relevant tasks; it is not identical to granting a runtime tool permission; the required outcomes are instead to delegate an independent code review without filling the main coding session with details and to continue a previously interrupted Claude Code work session.

D:

This alternative serves the separate goal 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; the required outcomes are instead to delegate an independent code review without filling the main coding session with details and to continue a previously interrupted Claude Code work session.

E:

This alternative serves the separate goal 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 required outcomes are instead to delegate an independent code review without filling the main coding session with details and to continue a previously interrupted Claude Code work session.

Question 319

An architecture decision for a privileged agent security review must support capture outcomes of agent command execution for security review and stop a coding agent from running commands that match a prohibited destructive pattern. Which option gives a complete rather than partial solution?

  1. Perform both steps: (1) Define fail-closed behavior for an enforcement hook and test failure paths; (2) enforce a pre-tool hook that denies the disallowed call.
  2. Perform both steps: (1) Define fail-closed behavior for an enforcement hook and test failure paths; (2) check resource-level authorization for every sensitive tool request.
  3. Perform both steps: (1) Record result metadata in a post-tool hook; (2) check resource-level authorization for every sensitive tool request.
  4. Perform both steps: (1) Record result metadata in a post-tool hook; (2) enforce a pre-tool hook that denies the disallowed call.

Explanation

Correct Answer

D:

The first change enables the team to capture outcomes of agent command execution for security review because post-execution hooks can log and inspect what happened; they are unsuitable as the sole barrier to a side effect that already occurred; the second addresses the separate need to stop a coding agent from running commands that match a prohibited destructive pattern because a deterministic hook checks the proposed command before execution, providing a stronger control than asking the agent to self-police.

Incorrect Answers

A:

The second step meets the requirement to stop a coding agent from running commands that match a prohibited destructive pattern because a deterministic hook checks the proposed command before execution, providing a stronger control than asking the agent to self-police; the first step would only help 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, leaving the separate requirement to capture outcomes of agent command execution for security review unmet.

B:

Both steps target different results—avoid accidentally allowing critical writes when a safety check crashes and stop an authenticated user from reading another tenant's records via a tool—because a guardrail that silently skips checks on errors is not an effective boundary; failures must prevent the protected action when policy requires it and proving identity does not grant access to every object; the backend must enforce tenant and resource permissions separately; the required capabilities are instead to capture outcomes of agent command execution for security review and to stop a coding agent from running commands that match a prohibited destructive pattern.

C:

The first step is appropriate for the requirement to capture outcomes of agent command execution for security review because post-execution hooks can log and inspect what happened; they are unsuitable as the sole barrier to a side effect that already occurred; the other step instead aims to stop an authenticated user from reading another tenant's records via a tool because proving identity does not grant access to every object; the backend must enforce tenant and resource permissions separately, which does not meet the second requirement to stop a coding agent from running commands that match a prohibited destructive pattern.

Question 320

A test for an incident involving failed Claude completions fails because it cannot find a request schema error instead of repeatedly changing the system prompt. Which change fixes the identified failure?

  1. Examine correlated request, tool, and transformation traces stage by stage.
  2. Trace the API credential and access permissions for the 401 or 403 response.
  3. Read the 400 error details and validate the request payload.
  4. Inspect stop_reason and output limits before changing the parser.

Explanation

Correct Answer

C:

To find a request schema error instead of repeatedly changing the system prompt in this incident involving failed Claude completions, an invalid API field or incompatible parameter cannot be fixed by natural-language instructions within the malformed request; the proposed implementation directly supplies that capability.

Incorrect Answers

A:

The proposed alternative helps to locate which component of a multi-tool answer introduced the wrong value because structured traces reveal whether retrieval, tool execution, postprocessing, or generation caused the defect; reading only the final text cannot isolate it; that mechanism does not meet the different requirement to find a request schema error instead of repeatedly changing the system prompt.

B:

The proposed alternative helps 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; that mechanism does not meet the different requirement to find a request schema error instead of repeatedly changing the system prompt.

D:

The proposed alternative helps to find why a JSON record unexpectedly ends midway through a field because a max_tokens stop indicates incomplete generation; the parser may be working correctly against a response that was never finished; that mechanism does not meet the different requirement to find a request schema error instead of repeatedly changing the system prompt.

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