Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 6

Question 50

Claude Code is working with an unfamiliar repository too large to read at once. An agent cannot tell where major services live in a large monorepo. Which control is most directly relevant?

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

Explanation

Correct Answer

D:

A file map supports targeted exploration and prevents repeatedly flooding context with unrelated source files. The relevant requirement is to create a concise navigation map, and this measure operates at that decision point.

Incorrect Answers

A:

A resumable scratchpad prevents rediscovery and reduces the risk of treating an old hypothesis as a verified fix; the alternative problem occurs when a lengthy investigation is paused before the final patch. The present scenario instead requires the system to create a concise navigation map.

B:

It would be appropriate when several subagents have inspected separate subsystems; a structured handoff preserves useful specialist evidence while limiting the coordinator's context load. Here, an agent cannot tell where major services live in a large monorepo, so a different mechanism is required.

C:

Focused context makes the debugging task tractable without deleting the ability to revisit excluded evidence; it protects against the case where search results span many modules but only two files are relevant to the current bug, rather than the failure described here: an agent cannot tell where major services live in a large monorepo.

E:

This choice is relevant when the working context is compacted after many exploratory reads; useful compaction shortens history while keeping the information needed to continue safely and verify the eventual result. Here, however, an agent cannot tell where major services live in a large monorepo.

Question 51

Claude Code is exploring a large monorepo for a maintenance task. An engineer knows a filename pattern but not its directory. How should the workflow be adjusted?

  1. Run the build or test using a controlled shell or task tool
  2. Use a file glob or filename discovery operation
  3. Use a file-read operation on the known path and relevant range
  4. Use a targeted content search with a bounded file scope

Explanation

Correct Answer

B:

Glob-style search identifies paths by name pattern without reading or executing each file in the repository. Here it serves the explicit need to locate candidate paths efficiently without adding a separate unverified inference.

Incorrect Answers

A:

Only an actual process invocation establishes compiler and test behavior; inspection tools cannot substitute for execution; the alternative problem occurs when a scripted maintenance operation must compile the project and observe test output. The present scenario instead requires the system to locate candidate paths efficiently.

C:

It would be appropriate when the engineer knows the exact configuration file and needs to inspect one section; direct reading preserves surrounding context when a file is already identified and avoids broad repository scans. Here, an engineer knows a filename pattern but not its directory, so a different mechanism is required.

D:

Text search finds matching symbols inside files; filename discovery alone cannot identify call sites; it protects against the case where an engineer needs every reference to a deprecated function in source text, rather than the failure described here: an engineer knows a filename pattern but not its directory.

Question 52

An engineering assistant investigates a large repository with long context demands. Repeatedly pasting whole files into one conversation makes later reasoning unreliable. Which implementation decision is the best fit?

  1. Provide a compact task-state recap with affected files, assumptions, and unresolved work
  2. Search for relevant symbols, read selected spans, and keep a file map
  3. Audit compacted state against a checklist of nonnegotiable rules
  4. Record source revision and file or symbol location alongside the claim

Explanation

Correct Answer

B:

Targeted context reduces distraction and helps Claude reason about the specific implementation rather than unrelated files.

Incorrect Answers

A:

Explicit state restoration prevents old side discussions from becoming accidental requirements for the current fix; it protects against the case where a developer returns to a bug after several unrelated investigations, rather than the failure described here: repeatedly pasting whole files into one conversation makes later reasoning unreliable.

C:

This choice is relevant when a critical security constraint disappears after context compaction; compaction should reduce volume, not silently remove guardrails necessary for safe continuation. Here, however, repeatedly pasting whole files into one conversation makes later reasoning unreliable.

D:

It would be appropriate when a summary says a function was removed but contains no commit or path evidence; code facts change over time; a reference to the inspected revision makes the assertion auditable. Here, repeatedly pasting whole files into one conversation makes later reasoning unreliable, so a different mechanism is required.

Question 53

A support assistant drafts responses that must meet measurable service standards. Reviewers disagree because the instruction merely says 'be helpful.' What should the architect specify in the design?

  1. Require policy lookup and verified eligibility fields before recommending a refund
  2. Specify acknowledgement, verified status, and a concrete authorized next action
  3. Specify required facts, tone limits, forbidden claims, and measurable completion checks
  4. Require policy identifier, effective date, and relevant decision rationale

Explanation

Correct Answer

C:

Explicit criteria guide generation and review; subjective adjectives alone do not define whether a reply satisfied the support policy.

Incorrect Answers

A:

The response contract should identify the evidence needed for a decision instead of rewarding confident language unsupported by account data; it protects against the case where an agent answers a refund inquiry without checking customer eligibility, rather than the failure described here: reviewers disagree because the instruction merely says 'be helpful.'.

B:

It would be appropriate when a support reply must acknowledge an issue and give the next permitted step; component-level criteria make the outcome testable while leaving room for natural wording within the approved service tone. Here, reviewers disagree because the instruction merely says 'be helpful.', so a different mechanism is required.

D:

This choice is relevant when a regulator requires every decision note to cite the policy version used; mandatory citation fields connect conclusions to the governing rule and make later audits possible when policies change. Here, however, reviewers disagree because the instruction merely says 'be helpful.'.

Question 54

A multi-agent developer assistant manages long-running refactoring tasks. Multiple workers create unrelated notes with no common file identifiers. Which change most directly addresses this condition?

  1. Pass a summary with linked files, decisions, and unresolved questions
  2. Mark stale notes superseded and record the new source revision
  3. Require entries keyed by task, file path, and evidence status
  4. Record the commit or diff identifier with each test result

Explanation

Correct Answer

C:

A consistent scratchpad makes cross-worker discoveries retrievable and reduces conflicting or orphaned notes. Here it serves the explicit need to adopt an indexable shared scratchpad without adding a separate unverified inference.

Incorrect Answers

A:

Raw transcripts contain irrelevant exploration and private context; scoped artifacts support collaboration more efficiently; the alternative problem occurs when a worker wants to reuse another agent's whole transcript as context. The present scenario instead requires the system to adopt an indexable shared scratchpad.

B:

This choice is relevant when an old scratchpad conflicts with an authoritative code search performed today; scratchpads are working memory, not authority; current repository evidence must win over obsolete summaries. Here, however, multiple workers create unrelated notes with no common file identifiers.

D:

A green status without a revision cannot establish validity after further edits; provenance is necessary for trustworthy continuation; it protects against the case where the context summary says 'tests passed' but omits which revision was tested, rather than the failure described here: multiple workers create unrelated notes with no common file identifiers.

Question 55

A document-processing agent decides whether to retry failed MCP calls. A service responds with a temporary capacity error and a recommended wait. What is the most defensible next step?

  1. Use an operation ID to query authoritative completion status
  2. Return a field-level constraint violation for the caller to repair
  3. Treat missing approval as a policy precondition requiring escalation
  4. Return a password-protection status rather than a fabricated empty conversion
  5. Use retry-aware error metadata with capped exponential backoff

Explanation

Correct Answer

E:

Temporary saturation may clear; paced retries can recover without causing a synchronized load spike. This matches the workflow goal to retry only after a bounded delay rather than introducing an unrelated action.

Incorrect Answers

A:

Retrying a financial mutation blindly can duplicate a transaction; checking durable state is safer than inferring failure from a timeout; the alternative problem occurs when a network timeout leaves the outcome of a state-changing operation uncertain. The present scenario instead requires the system to retry only after a bounded delay.

B:

This choice is relevant when a validator rejects a confidence value outside its supported range; a type or range error requires argument correction, not infrastructure backoff or an alternate source search. Here, however, a service responds with a temporary capacity error and a recommended wait.

C:

Approval cannot be manufactured by model retries; the workflow must obtain an authorized decision outside the model; it protects against the case where a tool reports that the user's approval is missing for a payout, rather than the failure described here: a service responds with a temporary capacity error and a recommended wait.

D:

It would be appropriate when an XML converter receives a password-protected file it cannot read; a clear input limitation enables the user to provide an accessible document through approved channels; repeated conversion is ineffective. Here, a service responds with a temporary capacity error and a recommended wait, so a different mechanism is required.

Question 56

An AI team uses a generator and reviewer for compliance reports. The writer evaluates its own output and overlooks repeated factual errors. Which approach satisfies the stated constraint?

  1. Preserve both positions and send the evidence to a human approver
  2. Keep review decisions in a distinct controlled stage or human gate
  3. Report the claim, evidence gap, and suggested bounded repair
  4. Use an independent reviewer prompt or agent grounded in source evidence

Explanation

Correct Answer

D:

A separate review role with explicit criteria can catch mistakes that a writer may repeat during unstructured self-review.

Incorrect Answers

A:

This choice is relevant when a dispute remains after the independent reviewer raises a high-risk concern; persistent disagreement is a signal for accountable judgment, not a reason to keep agents debating indefinitely. Here, however, the writer evaluates its own output and overlooks repeated factual errors.

B:

It would be appropriate when the writer is allowed to alter the reviewer verdict before publication; an independent check loses value if the generator can overwrite or ignore a failed result without accountability. Here, the writer evaluates its own output and overlooks repeated factual errors, so a different mechanism is required.

C:

Targeted feedback lets the writer fix the failing span without regenerating a document that already passed most checks; the alternative problem occurs when the reviewer identifies one unsupported paragraph among otherwise valid sections. The present scenario instead requires the system to create a separate verification pass.

Question 57

An application extracts structured fields from invoices with Claude. Free-form JSON text frequently contains trailing commentary that breaks parsing. Which mechanism should be selected?

  1. Model tax ID as optional or nullable according to downstream semantics
  2. Define an array of line-item objects with required amount and description fields
  3. Have Claude supply fields through a tool interface with a defined JSON Schema
  4. On a model that supports it, set tool_choice to tool with the required function name

Explanation

Correct Answer

C:

Tool-argument schemas provide a clearer machine-readable contract than merely asking for 'JSON only' in prose, though validation remains necessary.

Incorrect Answers

A:

It would be appropriate when some invoice records omit an optional tax identifier; a schema that requires nonexistent information encourages fabricated values; absence must be representable explicitly. Here, free-form JSON text frequently contains trailing commentary that breaks parsing, so a different mechanism is required.

B:

Nested arrays preserve line-item boundaries and permit validation of each element without relying on free-text delimiters; the alternative problem occurs when the system stores several line items for one invoice. The present scenario instead requires the system to use a constrained structured interface.

D:

A forced tool choice can enforce this output channel on compatible models; the application must still validate arguments and should use schema-constrained output when forced tool use is unavailable; it protects against the case where the deployment uses a Claude model and settings that support forcing a designated tool, and every extracted record must be submitted through that function, rather than the failure described here: free-form JSON text frequently contains trailing commentary that breaks parsing.

Question 58

A customer-support assistant must decide when to request human assistance. A request falls outside every published returns-policy category. What should be implemented at the appropriate boundary?

  1. Mark the policy gap and hand off to an authorized human
  2. Apply the verified rule within delegated permission and record the decision
  3. Pause account changes and request approved verification or support review
  4. Create a human-approval request with the verified amount and case evidence

Explanation

Correct Answer

A:

The model should not create a new rule for an unaddressed case; a documented gap requires accountable policy interpretation.

Incorrect Answers

B:

This choice is relevant when a support specialist finds the policy clear and customer eligibility already verified; human review should be focused on risk or uncertainty; routine well-defined cases can proceed under controlled authorization. Here, however, a request falls outside every published returns-policy category.

C:

Conflicting identity evidence makes a mutation unsafe; a fluent guessed choice is not a substitute for authentication; it protects against the case where the user provides conflicting account ownership details, rather than the failure described here: a request falls outside every published returns-policy category.

D:

It would be appropriate when a refund would exceed the support agent's approved limit; authority limits must be observed at execution; the model cannot increase its own financial permissions. Here, a request falls outside every published returns-policy category, so a different mechanism is required.

Question 59

A research pipeline asks Claude to emit machine-readable evidence records. A citation must include both source ID and publication date. Which option best meets the operational need?

  1. Allow a conditional locator such as a page or URL fragment as appropriate
  2. Check source IDs against the approved evidence registry after parsing
  3. Define source_id and publication_date as required schema properties
  4. Use an enum for allowed verification statuses

Explanation

Correct Answer

C:

Required fields expose missing citation metadata early instead of allowing evidence records that cannot later be traced.

Incorrect Answers

A:

It would be appropriate when some sources have no stable page number because they are HTML; the schema should reflect source-type differences; forcing a PDF page number for HTML would create meaningless metadata. Here, a citation must include both source ID and publication date, so a different mechanism is required.

B:

This choice is relevant when a tool-call payload satisfies syntax but cites an unknown document ID; JSON Schema checks shape and types, not whether a purported reference actually exists or supports the claim. Here, however, a citation must include both source ID and publication date.

D:

Enumerated values make downstream control flow predictable and prevent speculative status text from being treated as a recognized gate; it protects against the case where the model invents unrecognized status strings such as 'sort-of-verified', rather than the failure described here: a citation must include both source ID and publication date.

Question 60

A product team decomposes complex Claude tasks into subtasks. A customer email requires one classification label and no external data. What is the most appropriate design choice?

  1. Use an agent loop with step budgets and decisions conditioned on observations
  2. Implement a deterministic pipeline around Claude calls rather than a free-form planner
  3. Call Claude once with a precise classification instruction and output contract
  4. Extract, validate, calculate, then explain in that order
  5. Collect sources concurrently and synthesize once all results pass validation

Explanation

Correct Answer

C:

A single well-scoped task avoids unnecessary agent coordination when no independent dependencies or tools are needed.

Incorrect Answers

A:

Adaptive tool choice is useful for uncertain exploration, provided the host constrains possible actions and termination; the alternative problem occurs when the number of investigative steps cannot be known until initial evidence arrives. The present scenario instead requires the system to use the simplest sufficient execution pattern.

B:

This choice is relevant when a workflow repeats the same fixed five stages for every case; known stable stages are easier to test and audit as code; open-ended planning adds variance without meeting an unmet requirement. Here, however, a customer email requires one classification label and no external data.

D:

It would be appropriate when an analysis must validate extracted figures before calculating a ratio; the arithmetic is only meaningful after source values pass type and completeness checks; dependencies determine stage order. Here, a customer email requires one classification label and no external data, so a different mechanism is required.

E:

Independent lookups can run together, but writing should wait for the join to prevent omissions; it protects against the case where a report must combine three unrelated data sources before drafting, rather than the failure described here: a customer email requires one classification label and no external data.

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