Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 40

Question 390

An enterprise support assistant uses lengthy conversation and policy histories. A retrieved source conflicts with an earlier system-defined safety requirement. Which implementation decision is the best fit?

  1. Maintain a versioned decision ledger with approval status and source-turn references
  2. Treat retrieved text as evidence, not an override of higher-priority controls
  3. Fetch the latest verified address and label old entries as historical
  4. Restate a concise current policy excerpt near the final task instructions
  5. Retrieve the pertinent excerpt and carry its location into the answer

Explanation

Correct Answer

B:

External material can be adversarial or incorrect; long context must not promote a source document into policy authority.

Incorrect Answers

A:

A decision ledger tracks which resolution was authorized and which proposals were reversed, preventing a condensed conversation from treating tentative advice as binding; that assumes the support chat contains several competing ticket resolutions, including proposals that were later rejected. The item tests the different ability to maintain the instruction trust hierarchy.

C:

This control matters when a chat history includes several superseded account addresses; chronological conversation content can contain stale facts; an authoritative lookup prevents outdated messages from driving action. The required decision is different: a retrieved source conflicts with an earlier system-defined safety requirement.

D:

For the separate situation where the governing refund rule appears only once in the middle of a very long context, important details can be overlooked in long contexts; placing the controlling rule near use reduces lost-in-the-middle errors; the current task is to maintain the instruction trust hierarchy.

E:

Focused evidence helps the model judge a rule accurately and allows a reviewer to verify that the conclusion follows from the source; the alternative would make sense if an agent answers from a giant document without quoting the relevant passage. It does not directly fix the observed condition: a retrieved source conflicts with an earlier system-defined safety requirement.

Question 391

A code-inspection MCP server reports results and operational errors to Claude Code. The file parser crashes on an unsupported encoding. What should the architect specify in the design?

  1. Return an unsupported-encoding error and safe remediation choices
  2. Reject incomplete tool results before marking the inspection complete
  3. Report access denial with path scope but without privileged file contents
  4. Return a successful empty hit collection with the applied search scope

Explanation

Correct Answer

A:

An encoding-specific response directs the workflow to conversion or review instead of claiming the file contains no data.

Incorrect Answers

B:

Structural success is insufficient when required provenance fields are missing; validation must check the actual promised result schema; the alternative would make sense if a tool reports success but omits a required file hash. It does not directly fix the observed condition: the file parser crashes on an unsupported encoding.

C:

For the separate situation where the source file exists but cannot be read due to repository permissions, a permission-denied file is not equivalent to missing code; the operator may need an approved access adjustment; the current task is to classify the parser's input limitation.

D:

This control matters when a search query is valid and returns no matches in the target directory; zero results is useful evidence about a bounded search; presenting it as a tool failure would trigger unnecessary recovery. The required decision is different: the file parser crashes on an unsupported encoding.

Question 392

An agent runs a bounded support workflow with multiple optional tool calls. Support needs a record of why an agent issued each sensitive call. Which change most directly addresses this condition?

  1. Return a bounded handoff or clarification rather than forcing more autonomous turns
  2. Check cancellation state before dispatching queued tool work
  3. Log authorized tool names, request IDs, outcomes, and decision context
  4. Enforce a maximum tool-call or iteration budget and a fallback outcome

Explanation

Correct Answer

C:

Structured execution logs establish which external action occurred and on what evidence without assuming that model prose is a reliable audit record.

Incorrect Answers

A:

When prerequisites are missing, another iteration does not create evidence; explicit unresolved state protects correctness and enables review; the alternative would make sense if a response contains no tool calls but the task is still unresolved and requires human judgment. It does not directly fix the observed condition: support needs a record of why an agent issued each sensitive call.

B:

This control matters when a user cancels a request while a queued tool operation has not started; a cancellation flag in application state can block new operations; telling the model to stop is not sufficient for queued execution. The required decision is different: support needs a record of why an agent issued each sensitive call.

D:

For the separate situation where a tool-driven conversation keeps calling the same read-only endpoint without progress, without a host-side bound, repeatedly proposing the same action can consume resources without resolving the customer's issue; the current task is to retain an auditable tool decision trail.

Question 393

A Claude Code team shares MCP connections across several repositories. A new MCP server exposes a tool that should not be trusted automatically. What is the most defensible next step?

  1. Use a user-scoped MCP configuration for the reusable personal connection
  2. Store the connection in the supported project-scoped configuration
  3. Verify each effective user, project, and local configuration layer
  4. Audit server provenance, tool definitions, and permission prompts before enablement

Explanation

Correct Answer

D:

External servers can expose surprising side effects or injected descriptions; installation is not authorization for every capability.

Incorrect Answers

A:

This control matters when a developer uses the same safe personal reference server across many projects; user scope supports broadly useful tooling without copying the connection into every repository. The required decision is different: a new MCP server exposes a tool that should not be trusted automatically.

B:

For the separate situation where a repository requires a custom issue tracker MCP server not needed anywhere else, a project-level definition keeps repo-specific capabilities with the project instead of exposing them to unrelated workspaces; the current task is to review the capabilities before allowing execution.

C:

Effective tool availability can differ when local overrides or older definitions remain; inspecting the active scope resolves the discrepancy; the alternative would make sense if a team updates server configuration but different developers still see inconsistent tools. It does not directly fix the observed condition: a new MCP server exposes a tool that should not be trusted automatically.

Question 394

A code assistant integrates MCP tools maintained by different engineering groups. An API key appears in an MCP tool's result object. Which approach satisfies the stated constraint?

  1. Ignore instructions in tool metadata that conflict with higher-priority security policy
  2. Return an empty success result for legitimate no-match outcomes
  3. Redact secrets before returning tool content to Claude
  4. Use an enumerated argument value set and validate it on the server

Explanation

Correct Answer

C:

Tool responses become model context and may be quoted later; a restricted data boundary must filter credential material.

Incorrect Answers

A:

For the separate situation where one tool's description encourages bypassing repository review rules, descriptions are useful for discovery, not an authority source; a malicious tool can attempt prompt injection through its advertised text; the current task is to stop credential exposure through tool responses.

B:

No matching build is a valid result distinct from an operational error; mixing them causes unnecessary retries and escalation; that assumes a build lookup returns no match and callers mistake it for a crashed server. The item tests the different ability to stop credential exposure through tool responses.

D:

This control matters when a deployment tool accepts an environment name with many misspelled variants; schema constraints and server checks prevent an unintended target resulting from free-form environment strings. The required decision is different: an API key appears in an MCP tool's result object.

Question 395

A research pipeline emits structured literature reviews. The model keeps producing unsupported claims despite clean JSON. Which mechanism should be selected?

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

Explanation

Correct Answer

D:

Schema enforcement is not fact checking; a reviewer must compare meaning with evidence rather than regenerate well-formed guesses.

Incorrect Answers

A:

A number without its denominator can be misleading; semantic validation must check whether the claim is interpretable; that assumes a summarized study reports a percentage without a denominator. The item tests the different ability to introduce evidence-grounded review instead of more schema retries.

B:

A repair trail makes changes auditable and helps identify extraction defects hidden by a final clean object; the alternative would make sense if a publisher needs to know which assertions were repaired automatically. It does not directly fix the observed condition: the model keeps producing unsupported claims despite clean JSON.

C:

This control matters when a citation looks valid but links to a different paper title; a plausible-looking identifier can be factually wrong; checking external metadata tests referential truth beyond structure. The required decision is different: the model keeps producing unsupported claims despite clean JSON.

E:

For the separate situation where a citation record has a malformed DOI string, a targeted citation check catches formatting mistakes before publication while keeping evidence matching as a separate step; the current task is to introduce evidence-grounded review instead of more schema retries.

Question 396

A team uses Claude Code to implement a business-rule change. The generated patch contains unrelated formatting changes. What should be implemented at the appropriate boundary?

  1. Request the exact states and exception cases before coding
  2. Run the reproducer, inspect the cause, then apply a minimal change
  3. Update tests and plan based on the approved revised criteria
  4. Remove incidental changes and present a focused diff

Explanation

Correct Answer

D:

A small diff improves reviewability and makes unintended consequences easier to detect. This matches the workflow goal to reduce the patch to the intended behavior rather than introducing an unrelated action.

Incorrect Answers

A:

For the separate situation where a ticket uses the word 'active' without defining qualifying account states, an undefined business term can map to several implementations; clarification prevents a seemingly plausible but wrong patch; the current task is to reduce the patch to the intended behavior.

B:

This control matters when the team has a known bug with a reproducible failing test; a known failure gives objective feedback; starting with a broad rewrite increases unrelated regression risk. The required decision is different: the generated patch contains unrelated formatting changes.

C:

Continuing with an obsolete contract can produce polished but wrong code; the new requirements must govern validation; the alternative would make sense if a review discovers the acceptance criteria changed mid-implementation. It does not directly fix the observed condition: the generated patch contains unrelated formatting changes.

Question 397

A document classifier uses Claude to tag contracts with legal categories. A prompt example contains outdated policy thresholds. Which option best meets the operational need?

  1. Choose examples from different layouts and writing styles while keeping the rubric stable
  2. Show benign text that resembles the target but does not satisfy its criteria
  3. Replace obsolete examples and rerun evaluation on current cases
  4. Include labeled examples that explain how to break ties

Explanation

Correct Answer

C:

Few-shot content can encode policy mistakes as strongly as explicit instructions; stale exemplars should be treated as configuration debt.

Incorrect Answers

A:

This control matters when the training examples use only one document style, but production inputs vary; varied examples prevent the model from associating a label with incidental formatting instead of the clause meaning. The required decision is different: a prompt example contains outdated policy thresholds.

B:

Negative cases calibrate the decision boundary more directly than adding more obvious positive examples; that assumes reviewers identify recurring false positives from harmless boilerplate. The item tests the different ability to refresh demonstrations when underlying rules change.

D:

For the separate situation where a particular clause belongs to two related categories but reviewers expect one primary label, examples should show which category wins at overlapping boundaries rather than leaving Claude to infer an unstated hierarchy; the current task is to refresh demonstrations when underlying rules change.

Question 398

A developer uses Claude Code to change a risky authentication module. A proposed refactor requires an explicit human go/no-go. What is the most appropriate design choice?

  1. Perform the authorized edits and run the specified verification steps
  2. Use plan mode to inspect the codebase and propose steps without applying edits
  3. Present the plan and wait for authorization before enabling write operations
  4. Apply a direct targeted edit and verify the diff

Explanation

Correct Answer

C:

A design review requirement is not satisfied merely by generating a plan if the next stage can still commit changes autonomously.

Incorrect Answers

A:

Once scope and approval are established, indefinite re-planning adds little value; controlled execution produces testable results; the alternative would make sense if the approved implementation has already been planned and reviewed. It does not directly fix the observed condition: a proposed refactor requires an explicit human go/no-go.

B:

For the separate situation where the engineer wants a proposed design and affected-file list before any modifications, read-only planning supports review of dependencies and risk before the team authorizes source changes; the current task is to insert an approval checkpoint after planning.

D:

This control matters when a typo in a known Markdown file is approved for immediate correction; a small, well-scoped low-risk change can be executed directly when the user has approved it. The required decision is different: a proposed refactor requires an explicit human go/no-go.

Question 399

An engineering assistant writes migration plans for internal reviewers. A reviewer cares about data migration reversibility. Which response is most reliable?

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

Explanation

Correct Answer

C:

Rollback feasibility must be assessed during planning, not assumed after the assistant recommends irreversible changes.

Incorrect Answers

A:

Distinguishing established facts from hypotheses prevents speculative risk language from masquerading as a verified defect; that assumes a risk register mixes confirmed issues with guesses. The item tests the different ability to include rollback criteria in the design contract.

B:

This control matters when a rollout has a strict zero-downtime requirement; an explicit availability constraint changes feasible migration designs; it must be visible before the model proposes modifications. The required decision is different: a reviewer cares about data migration reversibility.

D:

For the separate situation where a plan says 'update services' without naming dependent components, an actionable architecture plan names components and checks rather than relying on unbounded generic tasks; the current task is to include rollback criteria in the design contract.

Question 400

Claude Code is working with an unfamiliar repository too large to read at once. A lengthy investigation is paused before the final patch. What should the team do to resolve this particular gap?

  1. Read the two implicated files and retain references to the other hits
  2. Retain constraints, decisions, evidence references, and remaining risks
  3. Store the current hypothesis, inspected files, pending test, and next action
  4. Merge file-indexed findings with status and unresolved issues
  5. Record directories, ownership boundaries, and key entry points before deep reading

Explanation

Correct Answer

C:

A resumable scratchpad prevents rediscovery and reduces the risk of treating an old hypothesis as a verified fix. For the stated case, the check is whether the system can checkpoint the exact unfinished work; this approach supplies that control.

Incorrect Answers

A:

This control matters when search results span many modules but only two files are relevant to the current bug; focused context makes the debugging task tractable without deleting the ability to revisit excluded evidence. The required decision is different: a lengthy investigation is paused before the final patch.

B:

Useful compaction shortens history while keeping the information needed to continue safely and verify the eventual result; the alternative would make sense if the working context is compacted after many exploratory reads. It does not directly fix the observed condition: a lengthy investigation is paused before the final patch.

D:

A structured handoff preserves useful specialist evidence while limiting the coordinator's context load; that assumes several subagents have inspected separate subsystems. The item tests the different ability to checkpoint the exact unfinished work.

E:

For the separate situation where an agent cannot tell where major services live in a large monorepo, a file map supports targeted exploration and prevents repeatedly flooding context with unrelated source files; the current task is to checkpoint the exact unfinished work.

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