Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 27

Question 260

A review job fetches the current branch after it has advanced, but its report is attached to an older commit. What should be fixed?

  1. Remove commit identifiers from all reports.
  2. Increase the model's output-token allowance.
  3. Bind checkout, analysis, and reported findings to the same intended commit identifier.
  4. Assume all commits on one branch have identical content.

Explanation

Correct Answer

C:

Revision consistency makes the findings reproducible and relevant to the exact code under review. Branch names can advance between queueing and checkout. Recording the exact analyzed commit prevents findings from being attached to code that was never inspected.

Incorrect Answers

A:

Hiding revision information makes the mismatch harder to detect.

B:

More response capacity does not align the inspected code with the reported revision.

D:

A branch can move between revisions with materially different code.

Question 261

A lengthy automated review is truncated in CI console output. What is the best delivery pattern?

  1. Increase prose repetition so important findings might appear twice.
  2. Report that no findings exist because the visible tail is empty.
  3. Discard the report and rely on the model to recall it next week.
  4. Store the complete report as a CI artifact and provide a concise summary with its location.

Explanation

Correct Answer

D:

An artifact preserves the full result beyond console limits while the summary supports quick inspection. The saved report should identify the analyzed revision and relevant run. Console truncation then becomes a presentation limit rather than loss of the review evidence itself.

Incorrect Answers

A:

Duplicating findings adds noise and does not provide reliable access to the complete output.

B:

A truncated display cannot establish the absence of findings in the full report.

C:

A future session is not a durable record of this run's findings.

Question 262

A CI test command is piped to tee to save its log. The test process fails, but tee succeeds and the shell job reports success. What should the workflow fix?

  1. Preserve and check the test process's exit status, using appropriate pipeline failure handling such as Bash pipefail.
  2. Delete failure lines from the saved log before CI reads it.
  3. Infer test success from the existence of the saved log file.
  4. Increase the model's review budget while preserving the same shell status handling.

Explanation

Correct Answer

A:

A pipeline may otherwise report only its last command's status. Capturing the test failure prevents a successful logging utility from masking a failed validation gate.

Incorrect Answers

B:

Editing diagnostic output conceals the failing tests and does not correct the pipeline's status semantics.

C:

A log can be created for either a successful or failed test run; its existence is not a passing signal.

D:

Additional analysis cannot change which process exit code the shell pipeline reports.

Question 263

A custom Claude Code subagent is intended only to inspect code and return findings. Which configuration best supports that role?

  1. Give it no file-reading capability but ask for line-specific findings.
  2. Rely exclusively on a name containing the word readonly.
  3. Grant all editing and deployment tools to avoid future configuration work.
  4. Expose the read and search tools it needs while excluding unnecessary mutation tools.

Explanation

Correct Answer

D:

A role-appropriate tool set supports evidence gathering without unrelated editing capabilities. The subagent's configured capabilities should match its investigative role. A read-only label in prose is weaker than excluding mutation tools from its actual tool set.

Incorrect Answers

A:

The subagent would lack the evidence needed to perform its assigned review.

B:

A descriptive name does not constrain the subagent's available operations.

C:

Broad capabilities exceed the defined analysis role and increase accidental action scope.

Question 264

A Claude Code MCP server starts but fails authentication. Its shared configuration references an API-key environment variable that is set in the developer's interactive shell but absent from the server process. What should be checked first?

  1. How the server process receives its configured environment, without printing the secret value.
  2. Whether every authentication failure proves the remote service is offline.
  3. Whether the API key can be placed in each user question as a temporary fix.
  4. Whether increasing the model's reasoning budget repairs the server's credentials.

Explanation

Correct Answer

A:

Authentication depends on the environment actually supplied to the launched process rather than a separate shell's state. Process-launch environments can differ from the terminal in which a developer tested a variable. Diagnose the configuration path without copying the credential into model-visible prompts or logs.

Incorrect Answers

B:

The observed missing credential is a local configuration issue and does not establish a service outage.

C:

Prompt text is not the appropriate server authentication channel and would expose the credential.

D:

Model effort cannot supply a missing variable in the external server process.

Question 265

A custom Claude Code subagent must have the content of an existing, model-invocable api-conventions skill available at startup. Which subagent frontmatter setting directly requests this preload?

  1. description: api-conventions
  2. skills: [api-conventions]
  3. name: api-conventions
  4. tools: api-conventions

Explanation

Correct Answer

B:

The skills field requests startup injection of the listed skill content. This gives the subagent the convention text before its first task action.

Incorrect Answers

A:

The description helps select a subagent for a task. Naming a skill there does not request injection of that skill's content at startup.

C:

The name identifies the subagent itself. Matching a skill's name does not connect its contents to the new agent's startup context.

D:

The tools field names callable tools, not skills to preload. A skill name in that field is not the documented preload mechanism.

Question 266

Claude Code needs to locate a function in a large repository. What is the most efficient first approach?

  1. Read every file in full before using any search.
  2. Guess the file path and immediately replace its contents.
  3. Generate a new function with the same name elsewhere.
  4. Use targeted symbol or text search, then inspect the relevant definition and callers.

Explanation

Correct Answer

D:

Search narrows the evidence needed to understand the function without loading unrelated files. After locating the definition, relevant callers reveal how the function is used and which invariants apply. A symbol hit alone should not trigger an edit before its context is understood.

Incorrect Answers

A:

A full repository dump wastes context when a targeted search can locate the symbol.

B:

Editing an unverified location risks changing the wrong implementation.

C:

Adding another definition does not locate or understand the existing one.

Question 267

A command failed because Claude Code ran it from the monorepo root instead of the service directory. What should the next attempt change?

  1. Install a new copy of the dependencies at the monorepo root before checking the service's scripts.
  2. Rerun the unchanged command from the root with a higher timeout.
  3. Set the intended working directory explicitly and confirm the service's command configuration.
  4. Move the service's package configuration into the root so the original command finds it.

Explanation

Correct Answer

C:

An explicit directory aligns relative paths and package scripts with the target service. Many build commands resolve files and dependencies relative to the working directory. An explicit execution context prevents a valid service command from being interpreted at the monorepo root.

Incorrect Answers

A:

Adding root dependencies may hide a path mistake and change unrelated project state. The next attempt should use the directory and configuration intended for the service command.

B:

More execution time does not resolve the known working-directory mismatch. The process would still look for the wrong local configuration or scripts.

D:

Relocating configuration changes the repository layout to accommodate an invocation error. Setting the intended working directory is the direct correction.

Question 268

A proposed fix imports a package that is not installed or declared. What should Claude Code check before keeping that import?

  1. Whether the generated code is syntactically indented.
  2. Check existing dependencies before adding a compatible declared package.
  3. Whether the package name sounds relevant.
  4. Whether the model has used the package in another conversation.

Explanation

Correct Answer

B:

Inspecting project dependencies avoids unsupported imports and unnecessary additions. An existing utility may already satisfy the requirement without expanding the dependency surface. If a new package is needed, its supported version and installation records must match the project.

Incorrect Answers

A:

Formatting cannot make an undeclared module available at runtime.

C:

A plausible name does not establish availability, compatibility, or project approval.

D:

Past use in another project provides no evidence about this repository's dependency set.

Question 269

A README example and the implementation disagree about a function's return type. The task is to update the example accurately. What evidence should guide the edit?

  1. A fabricated type that combines every possible result.
  2. The example with the most polished wording.
  3. Whichever type would make the example shorter.
  4. The current public contract, checked against executable behavior.

Explanation

Correct Answer

D:

Documentation should reflect the supported interface, so conflicting evidence needs a concrete check. A public contract and an observed implementation can themselves disagree, so the task may require identifying a defect rather than choosing prose by style. The example should not invent an unsupported interface.

Incorrect Answers

A:

Inventing an interface conceals the discrepancy and misleads callers.

B:

Writing style does not determine which return type the implementation supports.

C:

Brevity is not evidence of the API's actual contract.

Question 270

Claude Code edited a file and also changed a remote database through a tool. A user rewinds the local code checkpoint. What should the agent avoid assuming?

  1. That the remaining remote change should be reported when relevant.
  2. That rewinding local files automatically reversed the remote database mutation.
  3. That the local file state may have been restored according to the selected rewind operation.
  4. That the remote database can be queried to determine its current state.

Explanation

Correct Answer

B:

External side effects need their own verified reversal or reconciliation mechanism. Checkpointing concerns the state it actually tracks. A remote database needs its own authoritative readback and, if appropriate, an authorized compensating operation.

Incorrect Answers

A:

A material external change is part of the user's actual task state.

C:

Restoring tracked local edits is within checkpointing's intended scope.

D:

An authorized readback is an appropriate way to inspect the external effect.

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