Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 30

Question 290

Company policy requires all custom business tools to run inside an approved network segment. What is the appropriate next design choice?

  1. Use a local function dispatcher for private integrations
  2. Start a fresh isolated context for a subtask
  3. Run the harness in a controlled private environment
  4. Apply a sandbox to untrusted generated code

Explanation

Correct Answer

C:

A self-operated worker provides network and runtime control for executing the permitted functions; in production, the operator must maintain patching, scaling, observability, and worker isolation.

Incorrect Answers

A:

Application code executes a narrowly defined local tool and returns selected results to Claude; here, the controlling capability is that a self-operated worker provides network and runtime control for executing the permitted functions, but the API itself does not have arbitrary access to a customer's private internal network.

B:

A separate worker receives only the necessary files and reports a minimal result; the missing capability is more specific: a self-operated worker provides network and runtime control for executing the permitted functions; isolation only works when tool rights and returned text follow the same boundary.

D:

Isolation constrains filesystem, processes, secrets, and network access independently of model reasoning; the needed capability is different—a self-operated worker provides network and runtime control for executing the permitted functions—and a prompt claiming code is safe cannot replace runtime containment.

Question 291

A development tool launches a local MCP server and communicates through process input and output streams. How should this requirement be met?

  1. Use stdio for a local subprocess integration
  2. Use Streamable HTTP for remote MCP access
  3. Expose an MCP tool for an executable operation
  4. Protect remote MCP authentication and authorization

Explanation

Correct Answer

A:

Standard input and output suit tightly coupled local client-server interactions; the important constraint is that the server's standard output must remain protocol-clean instead of mixing human debug prints.

Incorrect Answers

B:

A network transport supports remote requests under explicit authentication and deployment controls; the requirement instead turns on this behavior: standard input and output suit tightly coupled local client-server interactions; opening an HTTP endpoint without access restrictions can expose powerful tools publicly.

C:

An MCP tool advertises callable functionality with input schemas and returns execution results; this cannot substitute for the behavior that standard input and output suit tightly coupled local client-server interactions, because tool calls still need explicit runtime permissions and backend validation.

D:

Credentials and resource-scoped checks ensure each caller sees only permitted tools and records; in contrast, standard input and output suit tightly coupled local client-server interactions, while protocol compatibility alone does not establish who the caller is.

Question 292

An assistant proposes changing an invoice amount but its field values exceed permitted bounds. What is the right implementation response?

  1. Validate generated action arguments server-side
  2. Require explicit confirmation for irreversible actions
  3. Verify tool permissions independently of text
  4. Record immutable provenance for sensitive decisions

Explanation

Correct Answer

A:

Backend checks enforce type, range, scope, and approval before any mutation; the model's fluent output is not an authoritative business validation result.

Incorrect Answers

B:

A confirmation gate makes the consequential action a separate authorized decision, whereas the requirement calls for another behavior: backend checks enforce type, range, scope, and approval before any mutation; interpreting ordinary conversational intent as final approval can cause unintended deletion.

C:

Server-side authorization checks bind each requested resource to the authenticated principal; however, the critical mechanism here is that backend checks enforce type, range, scope, and approval before any mutation, and a system-prompt ban on disclosure cannot replace real access control.

D:

Audit metadata preserves input references, applicable policy, actor, and versioned output; the requirement instead turns on this behavior: backend checks enforce type, range, scope, and approval before any mutation; a final answer without source history is not adequate for reconstructing the decision.

Question 293

The main coding assistant needs a security audit but should receive only a concise finding from a separate context. Which solution should the engineer adopt?

  1. Use a reusable Skill for repeatable know-how
  2. Use an interactive session for exploratory editing
  3. Use a specialized subagent for focused review
  4. Use a custom slash command for a named recurring action

Explanation

Correct Answer

C:

A subagent delegates bounded work with appropriate tools and returns results to the parent; the implementation must still recognize that the delegated agent should not inherit broader privileges than necessary.

Incorrect Answers

A:

A Skill packages reusable task guidance so the coding agent can apply a consistent procedure; in contrast, a subagent delegates bounded work with appropriate tools and returns results to the parent, while guidance does not grant access to a tool or enforce backend authorization.

B:

An interactive coding session supports iterative commands and review as context evolves; this addresses a separate concern, not the fact that a subagent delegates bounded work with appropriate tools and returns results to the parent; automating every command without supervision is unnecessary for a human-led investigation.

D:

A named command exposes reusable prompt-driven workflow entry points inside Claude Code; the missing capability is more specific: a subagent delegates bounded work with appropriate tools and returns results to the parent; a command definition still depends on the session's tool permissions and context.

Question 294

A research assistant needs to know when coverage is sufficient rather than searching indefinitely. Which option best fits the described workflow?

  1. Revise the plan after a failed prerequisite
  2. Avoid splitting when coordination overhead dominates
  3. Build a task plan with measurable completion criteria
  4. Use a fallback to a human decision-maker

Explanation

Correct Answer

C:

A plan with explicit deliverables and stop checks makes progress and completion observable; in production, planning quality does not replace runtime controls on tools and duration.

Incorrect Answers

A:

The agent re-evaluates dependencies and selects a safe next step based on the observed failure; here, the controlling capability is that a plan with explicit deliverables and stop checks makes progress and completion observable, but it should not bypass authorization or invent missing file contents.

B:

Keeping a bounded task local avoids orchestration latency and reconciliation errors; the missing capability is more specific: a plan with explicit deliverables and stop checks makes progress and completion observable; subagents should be justified by distinct work or isolation needs, not mere availability.

D:

Escalation transfers the ambiguous decision with relevant evidence and uncertainties; the needed capability is different—a plan with explicit deliverables and stop checks makes progress and completion observable—and generating a confident answer is not an acceptable replacement for the required human approval.

Question 295

The next investigative step must depend on what the previous tool returned, and paths are not enumerable in advance. Which option best fits the described workflow?

  1. Assign action limits in the agent harness
  2. Run an explicit fixed-stage workflow
  3. Use a model-directed agent loop
  4. Execute a parallel fan-out workflow

Explanation

Correct Answer

C:

Claude selects an action after observing each result, with the application enforcing iteration and permission limits, although the autonomy adds variability and demands hard budgets, logs, and a stopping policy.

Incorrect Answers

A:

Host-controlled counters and deadlines bound execution regardless of what Claude decides next; by comparison, the intended solution relies on this: claude selects an action after observing each result, with the application enforcing iteration and permission limits; prompt instructions alone cannot enforce a hard external quota.

B:

Application code controls the predetermined stage transitions and makes them reproducible in tests; the needed capability is different—claude selects an action after observing each result, with the application enforcing iteration and permission limits—and it cannot decide to introduce an unplanned research step after unexpected evidence appears.

D:

Independent branches can run at once and a reducer collects their outputs, lowering wall-clock time; however, the critical mechanism here is that claude selects an action after observing each result, with the application enforcing iteration and permission limits, and parallel execution is unsafe if branches mutate the same resource without coordination.

Question 296

A sample copied from a blog references a retired model name and requests now fail. Which approach targets the deciding condition?

  1. Check currently supported model IDs before deployment
  2. Run a canary rollout for a new model version
  3. Benchmark quality on the actual task
  4. Verify tool and thinking feature support

Explanation

Correct Answer

A:

The official model catalog identifies supported identifiers and feature availability; its operational boundary is that historical aliases should not be assumed to remain valid.

Incorrect Answers

B:

Limited exposure allows observation of quality, refusal, latency, and tool usage before full migration; this addresses a separate concern, not the fact that the official model catalog identifies supported identifiers and feature availability; a model upgrade can affect the entire user experience even when code is unchanged.

C:

A representative evaluation suite reveals which model handles the real format and edge cases; here, the controlling capability is that the official model catalog identifies supported identifiers and feature availability, but a marketing ranking cannot replace workload-specific measurement.

D:

Capability checks prevent incompatible parameter and feature combinations; the missing capability is more specific: the official model catalog identifies supported identifiers and feature availability; similar family names do not guarantee identical APIs or limits.

Question 297

A production service needs full control over the message history, tool dispatcher, stop conditions, and audit log. What is the most direct implementation choice?

  1. Build a custom Messages API loop
  2. Use a local function dispatcher for private integrations
  3. Use an external workflow engine
  4. Use the Claude Agent SDK

Explanation

Correct Answer

A:

A developer-managed loop explicitly handles tool_use, executes tools, returns tool_result, and continues as needed; the implementation must still recognize that every error, retry, and termination branch must be implemented by the application.

Incorrect Answers

B:

Application code executes a narrowly defined local tool and returns selected results to Claude; the missing capability is more specific: a developer-managed loop explicitly handles tool_use, executes tools, returns tool_result, and continues as needed; the API itself does not have arbitrary access to a customer's private internal network.

C:

A job engine coordinates state transitions and recovery outside the model's transient response; by comparison, the intended solution relies on this: a developer-managed loop explicitly handles tool_use, executes tools, returns tool_result, and continues as needed; it still requires model calls as activities rather than relying on prompts for transaction control.

D:

The SDK provides higher-level agent controls and integrations while the application supplies policies and oversight; the needed capability is different—a developer-managed loop explicitly handles tool_use, executes tools, returns tool_result, and continues as needed—and prebuilt orchestration does not remove responsibility for secrets and tool permission boundaries.

Question 298

One MCP integration is launched as a local subprocess in an editor, while another must serve authenticated clients over a network. Which TWO transports are suited to those different deployments? Choose TWO.

  1. Expose an MCP tool for an executable operation
  2. Protect remote MCP authentication and authorization
  3. Use stdio for a local subprocess integration
  4. Use Streamable HTTP for remote MCP access
  5. Use a client-side transport adapter

Explanation

Correct Answers

C:

Standard input and output suit tightly coupled local client-server interactions; a crucial qualification is that the server's standard output must remain protocol-clean instead of mixing human debug prints.

D:

A network transport supports remote requests under explicit authentication and deployment controls; in production, opening an HTTP endpoint without access restrictions can expose powerful tools publicly.

Incorrect Answers

A:

An MCP tool advertises callable functionality with input schemas and returns execution results; this addresses a separate concern, not the fact that standard input and output suit tightly coupled local client-server interactions; tool calls still need explicit runtime permissions and backend validation.

B:

Credentials and resource-scoped checks ensure each caller sees only permitted tools and records; here, the controlling capability is that standard input and output suit tightly coupled local client-server interactions, but protocol compatibility alone does not establish who the caller is.

E:

Transport abstraction lets the same protocol-level concepts operate over supported connection methods; the needed capability is different—standard input and output suit tightly coupled local client-server interactions—and a transport switch does not erase server deployment or security requirements.

Question 299

A monitoring agent has exhausted its approved retrieval budget without enough evidence for a diagnosis. What should be implemented to solve this problem?

  1. Escalate after repeated inconclusive tool results
  2. Ask a clarifying question when critical input is missing
  3. Retry a failed read with bounded backoff
  4. Finish with a user-facing answer after the last action

Explanation

Correct Answer

A:

Bounded fallback returns uncertainty and a trace for human follow-up rather than fabricating certainty; the important constraint is that extra iterations are not justified once the cost and confidence guardrails trigger.

Incorrect Answers

B:

Requesting an authorized scope prevents guesses from becoming unintended side effects; the requirement instead turns on this behavior: bounded fallback returns uncertainty and a trace for human follow-up rather than fabricating certainty; an internal plan cannot legitimately invent the missing tenancy information.

C:

Retrying after increasing delays can recover a transient read failure without changing external state; this cannot substitute for the behavior that bounded fallback returns uncertainty and a trace for human follow-up rather than fabricating certainty, because the caller must stop after a limited number of attempts and surface persistent failure.

D:

The model synthesizes a final response once no additional client tool request is needed; in contrast, bounded fallback returns uncertainty and a trace for human follow-up rather than fabricating certainty, while a raw tool result is not automatically a contextualized customer answer.

Question 300

A generated report stops mid-section solely because it reached the configured generation limit. Which of these alternatives serves the stated goal?

  1. Prune stale tool outputs from long histories
  2. Set an explicit max_tokens budget
  3. Continue after max_tokens when appropriate
  4. Resubmit relevant history for a follow-up

Explanation

Correct Answer

C:

The application treats a truncated completion differently from an intentional end and safely continues if possible; the implementation must still recognize that delivering a partial JSON document as complete would break downstream parsing.

Incorrect Answers

A:

Context management keeps decisive evidence while removing redundant or obsolete tool text; in contrast, the application treats a truncated completion differently from an intentional end and safely continues if possible, while blindly dropping the only record of a permission decision can change subsequent behavior.

B:

The request parameter limits output generation independently from the size of the prompt; this addresses a separate concern, not the fact that the application treats a truncated completion differently from an intentional end and safely continues if possible; a larger max_tokens value does not expand the model's input context window.

D:

The stateless API needs the prior context supplied again or reconstructed by the app; the missing capability is more specific: the application treats a truncated completion differently from an intentional end and safely continues if possible; separate requests do not automatically share private conversational state.

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