Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 72

Question 710

A release check on a Python Claude API backend found two unsatisfied conditions: the service must let a reviewer trace claims to the material actually consulted, and it must stay within input limits while keeping the most important conversation constraints. Which complete engineering plan satisfies both?

  1. Perform both steps: (1) Retain source identifiers with retrieved evidence and render supported citations; (2) enable required plugin or connector capabilities using least-privilege configuration.
  2. Perform both steps: (1) Delimit retrieved text as untrusted source material; (2) enable required plugin or connector capabilities using least-privilege configuration.
  3. Perform both steps: (1) Delimit retrieved text as untrusted source material; (2) apply a context retention policy and summarize earlier turns where suitable.
  4. Perform both steps: (1) Retain source identifiers with retrieved evidence and render supported citations; (2) apply a context retention policy and summarize earlier turns where suitable.

Explanation

Correct Answer

D:

The first change enables the team to let a reviewer trace claims to the material actually consulted because traceability requires linking claims to accessible source passages; fabricated URLs or unlabeled snippets do not provide verifiable provenance; the second addresses the separate need to stay within input limits while keeping the most important conversation constraints because selective history management preserves task-critical state; dropping random leading messages can erase consent or requirements.

Incorrect Answers

A:

The first step is appropriate for the requirement to let a reviewer trace claims to the material actually consulted because traceability requires linking claims to accessible source passages; fabricated URLs or unlabeled snippets do not provide verifiable provenance; the other step instead aims to expose external business tools only to workflows that actually need them because tools expand the application's action surface; indiscriminate activation increases risk and can confuse model selection among similar functions, which does not meet the second requirement to stay within input limits while keeping the most important conversation constraints.

B:

Both steps target different results—prevent a retrieved article's embedded commands from being treated as developer instructions and expose external business tools only to workflows that actually need them—because explicit trust boundaries help the model interpret documents as evidence rather than operational orders, and runtime controls must enforce privileged actions and tools expand the application's action surface; indiscriminate activation increases risk and can confuse model selection among similar functions; the required capabilities are instead to let a reviewer trace claims to the material actually consulted and to stay within input limits while keeping the most important conversation constraints.

C:

The second step meets the requirement to stay within input limits while keeping the most important conversation constraints because selective history management preserves task-critical state; dropping random leading messages can erase consent or requirements; the first step would only help to prevent a retrieved article's embedded commands from being treated as developer instructions because explicit trust boundaries help the model interpret documents as evidence rather than operational orders, and runtime controls must enforce privileged actions, leaving the separate requirement to let a reviewer trace claims to the material actually consulted unmet.

Question 711

A team maintaining a hosted agent rollout must meet 3 separate acceptance criteria: (1) restrict a development agent to a defined set of file and command capabilities; (2) stop iterative tool use when a success condition or resource budget is reached; (3) prevent specialists from producing incompatible artifacts for the orchestrator. Which three implementation actions directly satisfy the stated requirements? Choose THREE.

  1. Define structured inputs, expected outputs, and failure states for each delegate.
  2. Run independent branches concurrently and merge their outputs afterward.
  3. Use the Claude Agent SDK for the agent's execution harness.
  4. Persist and resume the agent session using supported state handling.
  5. Configure tool permissions rather than depending only on a prompt.
  6. Enforce explicit completion checks and a maximum iteration budget.

Explanation

Correct Answers

A:

The test to prevent specialists from producing incompatible artifacts for the orchestrator is met because explicit delegation contracts make multi-agent integration testable; free-form handoffs often obscure whether a child finished its assigned task; the remaining selections covers the separate need to restrict a development agent to a defined set of file and command capabilities and need to stop iterative tool use when a success condition or resource budget is reached.

E:

The test to restrict a development agent to a defined set of file and command capabilities is met because runtime permissions constrain what actions can run even if the model requests more access; prompt instructions alone are not a security boundary; the remaining selections covers the separate need to prevent specialists from producing incompatible artifacts for the orchestrator and need to stop iterative tool use when a success condition or resource budget is reached.

F:

The test to stop iterative tool use when a success condition or resource budget is reached is met because bounded loops avoid unending tool calls and uncontrolled spend while making incomplete work visible instead of silently continuing; the remaining selections covers the separate need to prevent specialists from producing incompatible artifacts for the orchestrator and need to restrict a development agent to a defined set of file and command capabilities.

Incorrect Answers

B:

The action is useful for the separate capability 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; however, the required test outcomes are to prevent specialists from producing incompatible artifacts for the orchestrator and to restrict a development agent to a defined set of file and command capabilities and to stop iterative tool use when a success condition or resource budget is reached.

C:

The action is useful for the separate capability to build a Claude-powered coding agent with an established runtime for tool orchestration because the SDK provides structured primitives for agent sessions and tool execution, reducing the custom loop infrastructure the application must maintain; however, the required test outcomes are to prevent specialists from producing incompatible artifacts for the orchestrator and to restrict a development agent to a defined set of file and command capabilities and to stop iterative tool use when a success condition or resource budget is reached.

D:

The action is useful for the separate capability to continue a paused multi-step task with its legitimate prior context because session continuity preserves relevant work across invocations; discarding session state can cause repeated tools or lost intermediate decisions; however, the required test outcomes are to prevent specialists from producing incompatible artifacts for the orchestrator and to restrict a development agent to a defined set of file and command capabilities and to stop iterative tool use when a success condition or resource budget is reached.

Question 712

The technical owner of a team-owned Claude Code workflow has to give Claude Code repository-specific build and test instructions for future sessions. A separate requirement is to introduce Claude Code to an unfamiliar codebase's conventions and commands. Which two-part remediation covers both?

  1. Perform both steps: (1) Use the supported session continuation or resume mechanism; (2) initialize project guidance and review the resulting claude.md content.
  2. Perform both steps: (1) Add concise project guidance to the appropriate CLAUDE.md file; (2) define policy in the appropriate managed or shared settings scope.
  3. Perform both steps: (1) Use the supported session continuation or resume mechanism; (2) define policy in the appropriate managed or shared settings scope.
  4. Perform both steps: (1) Add concise project guidance to the appropriate CLAUDE.md file; (2) initialize project guidance and review the resulting claude.md content.

Explanation

Correct Answer

D:

The first change enables the team 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 second addresses the separate need to introduce Claude Code to an unfamiliar codebase's conventions and commands because repository setup discovers useful project practices, but autogenerated guidance still needs a human check for accuracy and unnecessary detail.

Incorrect Answers

A:

The second step meets the requirement to introduce Claude Code to an unfamiliar codebase's conventions and commands because repository setup discovers useful project practices, but autogenerated guidance still needs a human check for accuracy and unnecessary detail; the first step would only help to continue a previously interrupted Claude Code work session because restoring session context reduces lost work while preserving continuity of prior tool activity; opening a new unrelated session does not do the same, leaving the separate requirement to give Claude Code repository-specific build and test instructions for future sessions unmet.

B:

The first step is appropriate for the requirement 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 other step instead aims 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, which does not meet the second requirement to introduce Claude Code to an unfamiliar codebase's conventions and commands.

C:

Both steps target different results—continue a previously interrupted Claude Code work session and enforce a team-approved permission policy despite local developer preferences—because restoring session context reduces lost work while preserving continuity of prior tool activity; opening a new unrelated session does not do the same and 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 capabilities are instead to give Claude Code repository-specific build and test instructions for future sessions and to introduce Claude Code to an unfamiliar codebase's conventions and commands.

Question 713

Two approval criteria are still open for a business workflow that calls Claude: the team must explain why a previously cached prompt is now charged as new input and distinguish API transport failures from a successfully returned model refusal. Which two implementation actions satisfy them? Choose TWO.

  1. Cache an eligible unchanged prefix with cache_control.
  2. Configure the max_tokens request parameter appropriately.
  3. Inspect changes to content and ordering before the cache breakpoint.
  4. Inspect HTTP status and structured error payload before interpreting model output.
  5. Inspect response.content blocks and metadata through the SDK's typed representation.

Explanation

Correct Answers

C:

To explain why a previously cached prompt is now charged as new input as part of a design that must also distinguish API transport failures from a successfully returned model refusal, cache reuse depends on matching prefix material; adding text to the front changes what is cached even if later text is unchanged; this is why the action is one of the two required selections.

D:

To distinguish API transport failures from a successfully returned model refusal as part of a design that must also explain why a previously cached prompt is now charged as new input, rate limits, authorization failures, and server errors happen at the request layer, unlike a normal response with a refusal stop reason; this is why the action is one of the two required selections.

Incorrect Answers

A:

This alternative serves the separate goal to reuse a long stable instruction or document prefix over repeated requests because prompt caching reduces repeated input processing when the prefix matches; changes before the cache boundary can prevent a hit; the required outcomes are instead to explain why a previously cached prompt is now charged as new input and to distinguish API transport failures from a successfully returned model refusal.

B:

This alternative serves the separate goal to cap the maximum output generated by a Messages API request because a generation limit controls maximum output length, not total cost or correctness; too low a value can truncate useful answers; the required outcomes are instead to explain why a previously cached prompt is now charged as new input and to distinguish API transport failures from a successfully returned model refusal.

E:

This alternative serves the separate goal to avoid assuming a structured SDK response is a plain string because the SDK exposes block types, stop reason, and usage separately; naïve string concatenation can lose tool requests or other content; the required outcomes are instead to explain why a previously cached prompt is now charged as new input and to distinguish API transport failures from a successfully returned model refusal.

Question 714

A team maintaining a token efficiency assessment must meet 2 separate acceptance criteria: (1) cut cost for a nightly noninteractive summarization workload; (2) prevent a tool_use block from being displayed as if it were natural-language text. Which two implementation actions directly satisfy the stated requirements? Choose TWO.

  1. Check each response block's type before extracting its fields.
  2. Prefer a maintained SDK when its typed methods and retry behavior fit the integration.
  3. Measure cache-write overhead against subsequent cache-read savings.
  4. Evaluate Message Batches instead of synchronous requests.
  5. Route simple requests to a cheaper evaluated tier and escalate difficult ones.

Explanation

Correct Answers

A:

The test to prevent a tool_use block from being displayed as if it were natural-language text is met because typed response objects can contain several block families; assuming every block has a text field causes logic errors; the remaining selection covers the separate need to cut cost for a nightly noninteractive summarization workload.

D:

The test to cut cost for a nightly noninteractive summarization workload is met because asynchronous batches can offer discounted processing when the application can tolerate deferred completion and result collection; the remaining selection covers the separate need to prevent a tool_use block from being displayed as if it were natural-language text.

Incorrect Answers

B:

The action is useful for the separate capability to choose between writing raw REST requests and using an official client library because sDKs wrap the same underlying service API and can simplify serialization, error types, and streaming without removing protocol constraints; however, the required test outcomes are to prevent a tool_use block from being displayed as if it were natural-language text and to cut cost for a nightly noninteractive summarization workload.

C:

The action is useful for the separate capability to reduce repeated cost for a stable lengthy policy prefix reused many times because cache writes may cost extra while hits are cheaper; the reuse frequency determines whether the optimization saves money; however, the required test outcomes are to prevent a tool_use block from being displayed as if it were natural-language text and to cut cost for a nightly noninteractive summarization workload.

E:

The action is useful for the separate capability to minimize average model cost when most requests are easy but some require deeper reasoning because cascading matches compute to task difficulty but requires a reliable escalation rule and measurement of misrouting; however, the required test outcomes are to prevent a tool_use block from being displayed as if it were natural-language text and to cut cost for a nightly noninteractive summarization workload.

Question 715

The team responsible for a tool-using incident triage agent needs to recover a multi-hour agent workflow after an infrastructure interruption. What is the most direct engineering change?

  1. Persist checkpoints for tool outputs and accepted decisions.
  2. Catch tool failures and report typed errors into the agent loop.
  3. Use the Claude Agent SDK for the agent's execution harness.
  4. Define and validate a typed schema at the handoff.

Explanation

Correct Answer

A:

To recover a multi-hour agent workflow after an infrastructure interruption in this tool-using incident triage agent, durable checkpoints let the orchestrator resume from a known stage rather than rerunning costly or side-effecting actions; the proposed implementation directly supplies that capability.

Incorrect Answers

B:

The proposed alternative helps to keep a failed tool from being mistaken for successful completion of an objective because explicit failure messages let the model or controller choose recovery; swallowing exceptions makes later actions rely on results that never existed; that mechanism does not meet the different requirement to recover a multi-hour agent workflow after an infrastructure interruption.

C:

The proposed alternative helps to build a Claude-powered coding agent with an established runtime for tool orchestration because the SDK provides structured primitives for agent sessions and tool execution, reducing the custom loop infrastructure the application must maintain; that mechanism does not meet the different requirement to recover a multi-hour agent workflow after an infrastructure interruption.

D:

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 recover a multi-hour agent workflow after an infrastructure interruption.

Question 716

To ship a TypeScript Claude integration, the application must use provider-hosted search or execution without implementing the tool body locally and map out-of-order batch outputs back to source records reliably. Which pair of listed actions is necessary for those stated capabilities? Choose TWO.

  1. Configure an available server-side tool with the required API definition.
  2. Assign and preserve unique custom identifiers for batched requests.
  3. Branch on stop_reason tool_use and dispatch the declared calls.
  4. Read the completed stream's usage and final message metadata.
  5. Use stop_sequences and inspect the returned stop_sequence field.

Explanation

Correct Answers

A:

To use provider-hosted search or execution without implementing the tool body locally as part of a design that must also map out-of-order batch outputs back to source records reliably, provider-executed tools are handled in Anthropic's environment and differ from client functions that the caller must run; this is why the action is one of the two required selections.

B:

To map out-of-order batch outputs back to source records reliably as part of a design that must also use provider-hosted search or execution without implementing the tool body locally, asynchronous results need stable correlation keys; relying on processing order can attach an output to the wrong input; this is why the action is one of the two required selections.

Incorrect Answers

C:

This alternative serves the separate goal to recognize that the model is asking the application to execute a client tool because a tool-use stop requires host-side handling and a follow-up message containing results, not presenting the call as final prose; the required outcomes are instead to use provider-hosted search or execution without implementing the tool body locally and to map out-of-order batch outputs back to source records reliably.

D:

This alternative serves the separate goal to account accurately for token consumption when a response is streamed because early text deltas do not necessarily carry authoritative final usage or stop information; accounting must wait for the appropriate terminal event; the required outcomes are instead to use provider-hosted search or execution without implementing the tool body locally and to map out-of-order batch outputs back to source records reliably.

E:

This alternative serves the separate goal to enforce a caller-defined textual boundary on generation because a stop sequence halts generation at a specified marker, distinct from natural completion and separate from a JSON schema validator; the required outcomes are instead to use provider-hosted search or execution without implementing the tool body locally and to map out-of-order batch outputs back to source records reliably.

Question 717

A release check on a shared Claude Code repository found two unsatisfied conditions: the service must make a reusable development procedure available to Claude Code with supporting files, and it must give Claude Code repository-specific build and test instructions for future sessions. Which complete engineering plan satisfies both?

  1. Perform both steps: (1) Place a path-targeted rule in the repository's rules configuration; (2) initialize project guidance and review the resulting claude.md content.
  2. Perform both steps: (1) Create a project skill with the documented skill metadata and content; (2) add concise project guidance to the appropriate claude.md file.
  3. Perform both steps: (1) Place a path-targeted rule in the repository's rules configuration; (2) add concise project guidance to the appropriate claude.md file.
  4. Perform both steps: (1) Create a project skill with the documented skill metadata and content; (2) initialize project guidance and review the resulting claude.md content.

Explanation

Correct Answer

B:

The first change enables the team 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 second addresses the separate need 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.

Incorrect Answers

A:

Both steps target different results—apply a coding convention only when Claude Code edits a particular path and introduce Claude Code to an unfamiliar codebase's conventions and commands—because scoped rules keep specialized instructions close to their applicable files, avoiding global instructions that conflict with other modules and repository setup discovers useful project practices, but autogenerated guidance still needs a human check for accuracy and unnecessary detail; the required capabilities are instead to make a reusable development procedure available to Claude Code with supporting files and to give Claude Code repository-specific build and test instructions for future sessions.

C:

The second step meets the requirement 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 first step would only help to apply a coding convention only when Claude Code edits a particular path because scoped rules keep specialized instructions close to their applicable files, avoiding global instructions that conflict with other modules, leaving the separate requirement to make a reusable development procedure available to Claude Code with supporting files unmet.

D:

The first step is appropriate for the requirement 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 other step instead aims to introduce Claude Code to an unfamiliar codebase's conventions and commands because repository setup discovers useful project practices, but autogenerated guidance still needs a human check for accuracy and unnecessary detail, which does not meet the second requirement to give Claude Code repository-specific build and test instructions for future sessions.

Question 718

An end-to-end test for an operations research assistant checks whether it can stop iterative tool use when a success condition or resource budget is reached and require human approval before an autonomous task crosses a high-impact decision boundary. Which plan should be approved to satisfy both checks?

  1. Perform both steps: (1) Persist and resume the agent session using supported state handling; (2) insert an approval gate before the high-risk action.
  2. Perform both steps: (1) Enforce explicit completion checks and a maximum iteration budget; (2) insert an approval gate before the high-risk action.
  3. Perform both steps: (1) Enforce explicit completion checks and a maximum iteration budget; (2) apply limits for steps, elapsed time, and token or tool expenditure.
  4. Perform both steps: (1) Persist and resume the agent session using supported state handling; (2) apply limits for steps, elapsed time, and token or tool expenditure.

Explanation

Correct Answer

B:

The first change enables the team to stop iterative tool use when a success condition or resource budget is reached because bounded loops avoid unending tool calls and uncontrolled spend while making incomplete work visible instead of silently continuing; the second addresses the separate need to require human approval before an autonomous task crosses a high-impact decision boundary because human-in-the-loop control interrupts execution at a defined boundary; an agent's natural-language assurance cannot substitute for authorization.

Incorrect Answers

A:

The second step meets the requirement to require human approval before an autonomous task crosses a high-impact decision boundary because human-in-the-loop control interrupts execution at a defined boundary; an agent's natural-language assurance cannot substitute for authorization; the first step would only help to continue a paused multi-step task with its legitimate prior context because session continuity preserves relevant work across invocations; discarding session state can cause repeated tools or lost intermediate decisions, leaving the separate requirement to stop iterative tool use when a success condition or resource budget is reached unmet.

C:

The first step is appropriate for the requirement to stop iterative tool use when a success condition or resource budget is reached because bounded loops avoid unending tool calls and uncontrolled spend while making incomplete work visible instead of silently continuing; the other step instead aims to impose predictable maximum cost and wall-clock exposure on autonomous execution because runtime budgets are external constraints on agent activity and must be enforced by the host, not inferred from a cooperative natural-language instruction, which does not meet the second requirement to require human approval before an autonomous task crosses a high-impact decision boundary.

D:

Both steps target different results—continue a paused multi-step task with its legitimate prior context and impose predictable maximum cost and wall-clock exposure on autonomous execution—because session continuity preserves relevant work across invocations; discarding session state can cause repeated tools or lost intermediate decisions and runtime budgets are external constraints on agent activity and must be enforced by the host, not inferred from a cooperative natural-language instruction; the required capabilities are instead to stop iterative tool use when a success condition or resource budget is reached and to require human approval before an autonomous task crosses a high-impact decision boundary.

Question 719

In a customer-facing AI feature, testing reveals that the system cannot continue a chat with relevant prior turns despite the API being stateless and cannot reuse a long stable instruction or document prefix over repeated requests. Select the two targeted actions that close these gaps. Choose TWO.

  1. Include the necessary conversation history in each Messages API call.
  2. Read the completed stream's usage and final message metadata.
  3. Attach the PDF using the supported document content form.
  4. Cache an eligible unchanged prefix with cache_control.
  5. Use stop_sequences and inspect the returned stop_sequence field.

Explanation

Correct Answers

A:

To continue a chat with relevant prior turns despite the API being stateless as part of a design that must also reuse a long stable instruction or document prefix over repeated requests, a response does not create server-side conversational memory by default; the client supplies retained turns and controls truncation; this is why the action is one of the two required selections.

D:

To reuse a long stable instruction or document prefix over repeated requests as part of a design that must also continue a chat with relevant prior turns despite the API being stateless, prompt caching reduces repeated input processing when the prefix matches; changes before the cache boundary can prevent a hit; this is why the action is one of the two required selections.

Incorrect Answers

B:

This alternative serves the separate goal to account accurately for token consumption when a response is streamed because early text deltas do not necessarily carry authoritative final usage or stop information; accounting must wait for the appropriate terminal event; the required outcomes are instead to continue a chat with relevant prior turns despite the API being stateless and to reuse a long stable instruction or document prefix over repeated requests.

C:

This alternative serves the separate goal to ask questions about information in a user-provided PDF through a supported document input because document analysis depends on actual supported content blocks and size limits; a local file path alone is not accessible to the remote model; the required outcomes are instead to continue a chat with relevant prior turns despite the API being stateless and to reuse a long stable instruction or document prefix over repeated requests.

E:

This alternative serves the separate goal to enforce a caller-defined textual boundary on generation because a stop sequence halts generation at a specified marker, distinct from natural completion and separate from a JSON schema validator; the required outcomes are instead to continue a chat with relevant prior turns despite the API being stateless and to reuse a long stable instruction or document prefix over repeated requests.

Question 720

To ship a context-rich document assistant, the application must handle a JSON parse failure without silently accepting partial data and reject model output missing a field that a downstream service requires. Which pair of listed actions is necessary for those stated capabilities? Choose TWO.

  1. Catch the parsing error and apply a bounded repair or retry path.
  2. Summarize prior decisions and preserve unresolved constraints.
  3. Validate parsed output against the application's declared schema.
  4. Apply domain-specific authorization and business rules after schema validation.
  5. Have a scoped subagent process sources and return a concise verified handoff.

Explanation

Correct Answers

A:

To handle a JSON parse failure without silently accepting partial data as part of a design that must also reject model output missing a field that a downstream service requires, defensive parsing makes failures explicit and limits reattempts; swallowing errors can propagate corrupt records; this is why the action is one of the two required selections.

C:

To reject model output missing a field that a downstream service requires as part of a design that must also handle a JSON parse failure without silently accepting partial data, parsing proves a string is legal JSON, but schema validation checks required keys, types, and structural constraints; this is why the action is one of the two required selections.

Incorrect Answers

B:

This alternative serves the separate goal to maintain the current task after older dialogue exceeds the active token budget because a faithful rolling summary carries forward objectives and commitments while avoiding a full-history replay on each call; the required outcomes are instead to handle a JSON parse failure without silently accepting partial data and to reject model output missing a field that a downstream service requires.

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 handle a JSON parse failure without silently accepting partial data and to reject model output missing a field that a downstream service requires.

E:

This alternative serves the separate goal to analyze a large source set without exposing every raw document to the parent agent because separate contexts let specialists work deeply while the coordinator sees only claims and references needed for the final decision; the required outcomes are instead to handle a JSON parse failure without silently accepting partial data and to reject model output missing a field that a downstream service requires.

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