Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 2

Question 10

A customer asks to change the destination of an active shipment. The carrier response states that the package has entered a restricted handoff stage and the available tool cannot change it. What should the assistant do next?

  1. Close the request as impossible without passing it to the available human exception process.
  2. Update the customer's saved address and treat that as redirecting the active package.
  3. Explain the tool limitation and escalate through the supported human process with the relevant shipment evidence.
  4. Retry the same change using exponential backoff for the rest of the handoff period.

Explanation

Correct Answer

C:

The requested change exceeds the available execution path. Escalation preserves the observed carrier state and allows an authorized operator to determine whether an exceptional remedy exists.

Incorrect Answers

A:

The tool limitation establishes that this automated path cannot act. It does not establish that the supported human process has no remedy, so closing the request prematurely discards that path.

B:

A profile update can affect future shipments without changing the carrier's active routing record. The restricted package still needs the carrier's supported exception process.

D:

Backoff addresses temporary availability problems. The response describes a business-state restriction, so repeated calls do not supply the unavailable redirection capability.

Question 11

A research worker returns a recommendation that conflicts with a current policy document. The recommendation includes no citations or dates. What is the best coordinator response before acting?

  1. Request the supporting source and resolve the conflict against the applicable current policy.
  2. Use whichever statement was received most recently in the conversation.
  3. Ask the same worker to repeat its recommendation with a confidence score.
  4. Request a vote from other workers that receive only the disputed recommendation.

Explanation

Correct Answer

A:

The coordinator needs provenance and scope to evaluate the discrepancy. Checking the governing policy prevents an unsupported worker summary from silently replacing authoritative evidence.

Incorrect Answers

B:

Message arrival time is not the policy's effective date. An older rule can be reported later than its replacement, so recency in the transcript cannot resolve authority.

C:

A second assertion and a confidence value still provide no source, applicability, or date. The conflict requires evidence that can be evaluated against the governing policy.

D:

Agreement among workers given the same unsupported input does not establish independent corroboration. They need the competing evidence and its scope to resolve the policy question.

Question 12

A repair agent reports that all tests passed. Its trace shows a proposed test command but no execution result. Which conclusion is justified?

  1. Static inspection of the patch is sufficient to confirm the reported test result.
  2. The success claim is unverified until an actual test result is obtained.
  3. The command proposal is enough to record the test suite as passed.
  4. The patch must be incorrect because no test result is visible.

Explanation

Correct Answer

B:

The trace distinguishes planning from observation. A trustworthy report should obtain the execution result or explicitly state that testing has not been verified.

Incorrect Answers

A:

Inspection can support a judgment about likely correctness, but it cannot prove a particular test run succeeded. The claim concerns execution evidence, not reviewer confidence.

C:

A proposed action demonstrates intent rather than execution. There may have been a launch failure, an interrupted process, or no invocation at all.

D:

Missing verification does not prove the patch is wrong. It means the reported validation cannot currently be substantiated.

Question 13

A generated report must satisfy a measurable rubric for coverage and citation support. Initial drafts are usually close, and specific reviewer feedback reliably improves them. Which pattern most directly supports repeated revision against that rubric?

  1. Alternate draft generation with rubric-based evaluation and targeted revision until acceptance or a defined limit.
  2. Apply the same fixed sequence of formatting transformations to every draft.
  3. Route the input once to a specialist selected by the report's subject.
  4. Generate several independent drafts and select the one with the greatest word count.

Explanation

Correct Answer

A:

The evaluator identifies concrete gaps and the optimizer revises the draft in response. Clear acceptance criteria and a bound keep improvement focused and prevent an indefinite revision loop.

Incorrect Answers

B:

A fixed transformation chain can standardize presentation. It does not adapt revisions to the particular factual and coverage defects identified by the rubric.

C:

Routing selects a suitable processing path. It does not itself create a feedback loop that evaluates a draft and directs a revision against the rubric.

D:

Parallel drafting offers alternatives, but length does not measure coverage or citation support. This selection rule does not use the reliable corrective feedback described in the stem.

Question 14

A booking assistant finds an available room, obtains customer approval, then waits ten minutes. Availability can change during that interval. What must happen before committing the reservation?

  1. Commit through an operation that checks current availability and handles conflicts explicitly.
  2. Store the room selection in the assistant's session and issue a confirmation before the booking response arrives.
  3. Refresh availability once, then write the reservation through an endpoint that does not detect intervening changes.
  4. Use the previously approved search result without another inventory check.

Explanation

Correct Answer

A:

The service can validate inventory at the transaction boundary. A conflict should trigger a clear alternative or renewed customer decision rather than a false success report.

Incorrect Answers

B:

Session state records the chosen room without securing supplier inventory. A customer-facing confirmation must wait for authoritative acceptance of the reservation.

C:

A separate read narrows the stale-data interval but leaves a race between checking and writing. The commit needs an atomic inventory check or an equivalent reservation mechanism.

D:

Approval captures the customer's intent, but the search result describes earlier inventory. It does not reserve the room against intervening bookings.

Question 15

Claude returns a client tool_use block naming lookup_invoice. The application has not yet called its invoice database. Which component must execute the lookup?

  1. The Anthropic service, because submitting a client-tool schema uploads its implementation.
  2. The application hosting the client-tool implementation.
  3. A second model request containing the same tool schema, without any database invocation.
  4. An MCP server selected automatically from the tool name, even though no such server is configured.

Explanation

Correct Answer

B:

The response contains a requested invocation rather than the database result. The host dispatches the named operation, validates its arguments, and supplies the resulting observation back to Claude.

Incorrect Answers

A:

A client-tool definition describes an interface; it does not upload the application's database function. Execution still belongs to the host that implements that interface.

C:

Another inference may request the tool again, but it cannot produce a verified database observation without execution. The host must obtain and return the actual result.

D:

MCP can provide tools when an integration exists. A client-tool name alone does not discover or authorize a server implementation.

Question 16

One assistant response requests two independent lookups with different tool_use IDs. The lookups complete in reverse order. How should their results be associated with the original requests?

  1. Set each tool_result's tool_use_id to the corresponding requested invocation ID.
  2. Generate a new tool-use identifier when each result arrives.
  3. Match each result to the tool name and ignore the invocation ID.
  4. Match results by their arrival time.

Explanation

Correct Answer

A:

The identifier links the observation to the exact call independently of completion order. Preserving it prevents one lookup's data from being interpreted as the answer to another.

Incorrect Answers

B:

A fresh identifier has no corresponding pending invocation in the prior assistant message. The result must retain the ID of the request it answers.

C:

A name identifies an operation type, not a particular call. Concurrent calls to the same tool can use different inputs, so the invocation ID is needed for unambiguous correlation.

D:

Execution timing can differ because of network or backend latency. Arrival order therefore does not reliably identify the requested operation.

Question 17

A customer lookup runs but fails because the supplied account ID does not exist. The host wants Claude to distinguish this failure from a successful lookup returning customer data. Which response is appropriate?

  1. Return the account error as an unrelated assistant text message rather than a tool result.
  2. Abort the entire conversation as a transport failure without returning a result for the call.
  3. Return a tool_result for the original call with is_error set to true and a useful failure description.
  4. Return a successful result containing an empty object and no error indication.

Explanation

Correct Answer

C:

The error marker distinguishes failed execution from a normal observation. A concise description such as an unknown account ID helps Claude ask for correction or choose a supported recovery step.

Incorrect Answers

A:

That message does not satisfy the pending tool invocation. The failure belongs in the correlated result so the assistant can associate it with the attempted lookup.

B:

The API exchange itself can be healthy while the application lookup fails. Returning a tool error preserves the invocation record and lets the assistant recover within the conversation.

D:

An empty success value obscures the distinction between absent optional data and failed lookup execution. The caller needs a failure signal to choose a corrective action.

Question 18

An assistant confuses search_orders, which returns matching summaries, with get_order, which retrieves one order by exact ID. Both tools have the description order lookup. What change most directly addresses this ambiguity?

  1. Keep identical descriptions and rely on the parameter property types to convey all selection rules.
  2. Describe each tool's selection criteria, expected inputs, and returned information distinctly.
  3. Return longer order descriptions while leaving tool definitions unchanged.
  4. Always force get_order, even when the user supplies only a customer name.

Explanation

Correct Answer

B:

The assistant needs to know when an exact identifier is required and when a query is appropriate. Clear behavioral contracts reduce selection errors before execution begins.

Incorrect Answers

A:

Types can constrain a value's shape without explaining whether the operation searches or retrieves a known record. Distinct operational descriptions are still needed.

C:

Richer results can help after the correct call, but they do not clarify which operation to choose initially. The selection ambiguity exists in the interface description.

D:

The exact-ID retrieval tool cannot replace discovery when no order identifier is known. Forcing one tool removes a necessary branch rather than clarifying the choice.

Question 19

A tool accepts a syntactically valid JSON object containing an account ID. The current user is not authorized to access that account. Which additional control is required?

  1. Validate that the account ID matches the required string pattern.
  2. Check account access against the authenticated user's permissions in the backend.
  3. Allow access whenever the account exists in the database.
  4. Authorize access using an owner_user_id supplied by the model in the same request.

Explanation

Correct Answer

B:

Schema validation establishes the argument's shape, not the caller's entitlement. The service must independently bind the requested account to an authorized identity before returning its data.

Incorrect Answers

A:

An identifier can be syntactically valid and still belong to another tenant. Pattern validation does not decide whether this authenticated user may access the record.

C:

Existence checks distinguish valid objects from unknown ones, but valid objects can belong to other users. Record existence is not an access grant.

D:

Model-supplied identity claims are untrusted input. The backend must derive the caller's identity from its authenticated context and check the requested account against it.

Question 20

An MCP server must expose a policy document as URI-addressable context and provide an explicit operation that submits a support ticket. Which pair of server primitives directly matches those requirements?

  1. Expose the policy as a resource and ticket submission as a tool.
  2. Expose both capabilities as resources whose reads encode the requested actions.
  3. Expose the policy as a prompt template and ticket creation as a resource read.
  4. Expose both capabilities as prompt templates and have the client infer which backend action to execute.

Explanation

Correct Answer

A:

A resource represents retrievable context, while a tool exposes an operation the application may invoke. This separation makes reading policy and changing ticket state distinct capabilities.

Incorrect Answers

B:

The document fits a resource, but ticket creation is an explicit state-changing invocation. Hiding it behind a read makes the action boundary unclear and does not match the specified primitive roles.

C:

Prompt templates describe reusable interactions rather than a URI-addressable document. A resource read also does not express the requested explicit ticket-submission operation.

D:

Prompts may guide a workflow, but a template is not itself the backend operation. This design lacks the direct context and action contracts requested in the stem.

Claude Certified Architect - Foundations (CCAR-F) (CCA-F) Practice Tests