Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 3

Question 20

A development agent explores competing fixes across multiple sessions. A developer needs to continue yesterday's unfinished refactor. Which option best meets the operational need?

  1. Fork a common session context for each experiment
  2. Rerun relevant tests on the current revision before accepting success
  3. Use separate worktrees or sandboxes for mutable branch experiments
  4. Preserve goals, constraints, decisions, open issues, and source references while discarding noise
  5. Resume the relevant session and inspect the recorded pending work

Explanation

Correct Answer

E:

Continuing the actual conversation preserves intent and decisions, although the current code must still be validated.

Incorrect Answers

A:

Branching keeps baseline assumptions comparable, whereas serially editing one conversation blends conclusions; it protects against the case where two proposed fixes need separate experiments from the same baseline, rather than the failure described here: a developer needs to continue yesterday's unfinished refactor.

B:

Tests describe a specific code state; a prior green result cannot establish correctness for a changed repository; the alternative problem occurs when a resumed worker reports success based on a cached test from last week. The present scenario instead requires the system to recover the existing task trajectory.

C:

It would be appropriate when one experiment altered local files that another experiment must not see; forked dialogue history alone does not isolate disk writes; workspace separation prevents cross-experiment contamination. Here, a developer needs to continue yesterday's unfinished refactor, so a different mechanism is required.

D:

This choice is relevant when a session has grown so large that prior assumptions hide the active requirement; structured compaction retains what is needed for continuation and reduces irrelevant history without implying forgotten facts are still available. Here, however, a developer needs to continue yesterday's unfinished refactor.

Question 21

A multi-source intelligence assistant must preserve provenance in its conclusions. Two sources give different values for the same metric on different dates. What is the most appropriate design choice?

  1. Record file type, version, denominator, and calculation method for each source
  2. Store each value alongside publication and observation dates
  3. Record the direct document locator and relevant passage reference
  4. Label the date unknown and qualify conclusions that depend on recency

Explanation

Correct Answer

B:

Different dates may describe genuine changes rather than factual contradiction; temporal provenance enables accurate interpretation.

Incorrect Answers

A:

A discrepancy may arise from different populations or transformations; provenance details are needed before choosing a value; the alternative problem occurs when a CSV report and a narrative article disagree about a count. The present scenario instead requires the system to retain source timestamps with each quantitative claim.

C:

A generic homepage does not let reviewers trace the asserted evidence; precise provenance makes claims auditable; it protects against the case where a summary cites the homepage of a website but not the page containing a key claim, rather than the failure described here: two sources give different values for the same metric on different dates.

D:

This choice is relevant when a retrieved document lacks a reliable publication date; a missing timestamp cannot be repaired by inference from a filename or crawl date; time-sensitive claims need verified dates. Here, however, two sources give different values for the same metric on different dates.

Question 22

A research pipeline emits structured literature reviews. A citation record has a malformed DOI string. Which response is most reliable?

  1. Check the identifier format and request a corrected value when supported by the source
  2. Require the base population or flag the statistic as incomplete
  3. Record initial failures, corrected fields, and validator outcomes
  4. Check claims against retrieved sources and reject unsupported assertions

Explanation

Correct Answer

A:

A targeted citation check catches formatting mistakes before publication while keeping evidence matching as a separate step.

Incorrect Answers

B:

It would be appropriate when a summarized study reports a percentage without a denominator; a number without its denominator can be misleading; semantic validation must check whether the claim is interpretable. Here, a citation record has a malformed DOI string, so a different mechanism is required.

C:

This choice is relevant when a publisher needs to know which assertions were repaired automatically; a repair trail makes changes auditable and helps identify extraction defects hidden by a final clean object. Here, however, a citation record has a malformed DOI string.

D:

Schema enforcement is not fact checking; a reviewer must compare meaning with evidence rather than regenerate well-formed guesses; the alternative problem occurs when the model keeps producing unsupported claims despite clean JSON. The present scenario instead requires the system to validate citation identifiers syntactically.

Question 23

A research agent must explore conflicting hypotheses with limited resources. The search space has a fixed list of ten known documents. What should the team do to resolve this particular gap?

  1. Allow a bounded agent loop to generate follow-ups from validated findings
  2. Enumerate documents and allocate bounded extraction tasks to each
  3. Track completed task intents and reuse valid existing findings
  4. Assign explicit confirmation and challenge tasks before the final synthesis

Explanation

Correct Answer

B:

A finite inventory can be processed predictably without an autonomous planner deciding what to look at next. Here it serves the explicit need to use deterministic coverage over a known set without adding a separate unverified inference.

Incorrect Answers

A:

Uncertain discovery benefits from adaptive next steps; the budget and tools still require host control; it protects against the case where an initial article may reveal unknown organizations that require follow-up queries, rather than the failure described here: the search space has a fixed list of ten known documents.

C:

Deduplicating work prevents an apparent multi-agent design from simply repeating the same investigation; the alternative problem occurs when the coordinator keeps asking new workers for the same question. The present scenario instead requires the system to use deterministic coverage over a known set.

D:

It would be appropriate when an answer requires independently checking both supportive and contradictory evidence; separate search objectives reduce confirmation bias and give the coordinator grounds to compare conflicting material. Here, the search space has a fixed list of ten known documents, so a different mechanism is required.

Question 24

A code analysis service accepts Claude-generated review findings. A file path can be missing for repository-wide design findings. Which control is most directly relevant?

  1. Use a typed finding schema that permits a null path only for global findings
  2. Declare an enum aligned to the triage system
  3. Verify referenced symbols and locations against the actual code revision
  4. Define whether extra keys are permitted and enforce that contract

Explanation

Correct Answer

A:

Conditional optional fields prevent fabricated paths while retaining precise locations for line-specific defects. Here it serves the explicit need to model optional locations intentionally without adding a separate unverified inference.

Incorrect Answers

B:

Controlled labels avoid synonyms that break deterministic routing and reporting; it protects against the case where the pipeline needs severities that match its existing triage labels, rather than the failure described here: a file path can be missing for repository-wide design findings.

C:

This choice is relevant when a review is structurally valid but flags a function absent from the repository; a well-formed schema cannot prove an architectural claim is true; source checks are still necessary. Here, however, a file path can be missing for repository-wide design findings.

D:

Unknown fields should have a deliberate policy; silent acceptance can hide model drift, while arbitrary rejection can break legitimate extensibility; the alternative problem occurs when the downstream parser rejects a valid JSON object because an unexpected property appears. The present scenario instead requires the system to model optional locations intentionally.

Question 25

An MCP client connects to business-data tools provided by third-party servers. A server's advertised description says it may ignore the user's data-access restrictions. How should the workflow be adjusted?

  1. Prevent retrieved text from authorizing tool calls or changing instruction hierarchy
  2. Treat tool annotations as hints and evaluate real side effects
  3. Enforce caller authorization in the client and backend regardless of tool descriptions
  4. Mark pending operations unknown and reconnect or reconcile through supported state
  5. Bind identity and allowed resources at the server execution boundary

Explanation

Correct Answer

C:

Server metadata is untrusted discovery data; access checks must remain enforceable even if metadata is malicious or misleading.

Incorrect Answers

A:

This choice is relevant when an MCP server returns a copied webpage containing hostile 'system' instructions; prompt-injection material may arrive through normal tool results, so the agent's higher-priority policies and allowlists remain in force. Here, however, a server's advertised description says it may ignore the user's data-access restrictions.

B:

Capability annotations improve discovery but are not a security boundary or proof that code has no writes; it protects against the case where a tool is marked read-only in descriptive metadata but its implementation writes audit records, rather than the failure described here: a server's advertised description says it may ignore the user's data-access restrictions.

D:

It would be appropriate when a remote MCP server drops connection during a multi-step task; transport failure does not establish that a side effect failed; the client must track request outcomes and avoid blind mutation retries. Here, a server's advertised description says it may ignore the user's data-access restrictions, so a different mechanism is required.

E:

A shared server must not conflate clients; prompt context and tool-list filtering alone are insufficient to isolate data access; the alternative problem occurs when multiple users connect to the same server with different data permissions. The present scenario instead requires the system to deny server metadata authority over access policy.

Question 26

A customer-support service drives Claude through an application-owned tool loop. The API replies with stop_reason=tool_use and an approved lookup request. Which implementation decision is the best fit?

  1. Return a tool_result marked as an error with a sanitized diagnostic
  2. Execute approved calls and return one result correlated to each tool-use identifier
  3. Deliver the assistant's final content without dispatching another tool
  4. Execute the requested tool and send its correlated tool_result before asking Claude to continue

Explanation

Correct Answer

D:

A tool-use stop transfers execution to the host application; returning the result in a subsequent message lets Claude incorporate actual data rather than guess.

Incorrect Answers

A:

A structured failed result preserves the agent loop and lets the model choose recovery or escalation; fabricating successful data would hide the failure; the alternative problem occurs when a tool request fails because a backend times out. The present scenario instead requires the system to finish the pending model-requested lookup.

B:

It would be appropriate when a single assistant message contains two independent tool_use blocks; tool-use IDs associate responses with particular requests; matching both blocks prevents a result from being mistaken for another operation. Here, the API replies with stop_reason=tool_use and an approved lookup request, so a different mechanism is required.

C:

An end-of-turn stop signals that the model has finished this turn; a fresh tool invocation would be an unauthorized extra operation rather than required orchestration; it protects against the case where the API replies with stop_reason=end_turn and no pending tool requests, rather than the failure described here: the API replies with stop_reason=tool_use and an approved lookup request.

Question 27

A Claude Code repository is organized into multiple services with different local rules. The root instructions list every service's commands and overwhelm simple tasks. What should the architect specify in the design?

  1. Delete or archive obsolete operational details from shared instructions
  2. Keep common conventions at root and isolate service-specific guidance
  3. Use a supported CLAUDE.md import or concise reference to the maintained document
  4. Update the path or import so Claude Code can load the maintained guidance

Explanation

Correct Answer

B:

Layered instructions reduce irrelevant context while preserving shared policies that apply across the repository. Here it serves the explicit need to move specialized rules closer to their scope without adding a separate unverified inference.

Incorrect Answers

A:

Outdated incident context can mislead future tasks; persistent memory should contain durable conventions and verified references; the alternative problem occurs when the team's root file contains a temporary incident note from last month. The present scenario instead requires the system to move specialized rules closer to their scope.

C:

Importing maintained guidance avoids parallel copies that drift when the release process changes; it protects against the case where a release policy document already exists and should not be copied into multiple instruction files, rather than the failure described here: the root instructions list every service's commands and overwhelm simple tasks.

D:

It would be appropriate when an imported policy file is no longer present after a directory move; a prompt reminder cannot replace a broken import; resolving the file reference restores the policy's discoverability. Here, the root instructions list every service's commands and overwhelm simple tasks, so a different mechanism is required.

Question 28

A code assistant must choose among repository inspection and execution tools. An engineer only asks which files define an interface by name. Which change most directly addresses this condition?

  1. Use repository search or symbol discovery on the relevant scope
  2. Invoke the approved test runner and inspect its exit status and output
  3. Allow file reads and searches while withholding file-edit tools
  4. Restrict edits and shell working directory to the authorized path

Explanation

Correct Answer

A:

Name-location tasks need index or text inspection; shell execution is unnecessary and may introduce side effects. Here it serves the explicit need to prefer a targeted search rather than running code without adding a separate unverified inference.

Incorrect Answers

B:

A model-generated prediction cannot verify runtime behavior; the observed test result is the relevant evidence; it protects against the case where the team asks whether the current test suite passes after a patch, rather than the failure described here: an engineer only asks which files define an interface by name.

C:

It would be appropriate when a configuration change must be proposed but not written until review; a planning phase should not silently mutate the workspace before the human approves the proposed patch. Here, an engineer only asks which files define an interface by name, so a different mechanism is required.

D:

Worker-specific file boundaries reduce the risk that a narrowly assigned fix changes unrelated services; the alternative problem occurs when a test-fix worker has permission to edit only a small package. The present scenario instead requires the system to prefer a targeted search rather than running code.

Question 29

An engineering assistant writes migration plans for internal reviewers. A plan says 'update services' without naming dependent components. What is the most defensible next step?

  1. Request checkpoints, reversal conditions, and limitations of restoring state
  2. Require affected service names, dependency edges, and verification steps
  3. Require each risk to be marked observed, inferred, or unknown with supporting evidence
  4. Require backward compatibility, staged rollout, and rollback considerations

Explanation

Correct Answer

B:

An actionable architecture plan names components and checks rather than relying on unbounded generic tasks. For the stated case, the check is whether the system can make scope completeness testable; this approach supplies that control.

Incorrect Answers

A:

Rollback feasibility must be assessed during planning, not assumed after the assistant recommends irreversible changes; the alternative problem occurs when a reviewer cares about data migration reversibility. The present scenario instead requires the system to make scope completeness testable.

C:

It would be appropriate when a risk register mixes confirmed issues with guesses; distinguishing established facts from hypotheses prevents speculative risk language from masquerading as a verified defect. Here, a plan says 'update services' without naming dependent components, so a different mechanism is required.

D:

An explicit availability constraint changes feasible migration designs; it must be visible before the model proposes modifications; it protects against the case where a rollout has a strict zero-downtime requirement, rather than the failure described here: a plan says 'update services' without naming dependent components.

Question 30

A team wants repeatable Claude Code workflows for contributors. Developers repeatedly invoke a fixed changelog-generation routine by name. Which approach satisfies the stated constraint?

  1. Design the command to accept and validate a module argument
  2. Put the command in the user's personal Claude Code command scope
  3. Package the instructions and supporting resources as a Claude Code Skill
  4. Create a project-scoped custom command with clear inputs and instructions
  5. Keep a concise skill overview and reference supplemental files as needed

Explanation

Correct Answer

D:

A named command makes an explicit routine easy to invoke consistently without embedding the same instructions in every chat.

Incorrect Answers

A:

Parameterized workflows remain reusable while tying file operations to a caller-specified scope instead of assuming a default; the alternative problem occurs when a command needs a target module supplied by the developer. The present scenario instead requires the system to package a deliberate reusable command.

B:

It would be appropriate when an individual uses a private note-cleanup command in many projects; user-level commands support individual workflow shortcuts without adding unapproved tasks to team configuration. Here, developers repeatedly invoke a fixed changelog-generation routine by name, so a different mechanism is required.

C:

Skills can encapsulate specialized procedures with descriptions that support appropriate discovery and reuse; it protects against the case where a repository has a capability that Claude should discover when task context matches, rather than the failure described here: developers repeatedly invoke a fixed changelog-generation routine by name.

E:

This choice is relevant when a skill includes several bulky examples rarely needed for selection; progressive disclosure preserves context while still allowing detailed procedures to be consulted for the actual task. Here, however, developers repeatedly invoke a fixed changelog-generation routine by name.

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