Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 17

Question 160

One MCP service is shared across tenants, and a tool accepts a caller-provided tenant ID. What must the backend enforce?

  1. Derive or verify tenant authorization on the server for every tool request
  2. Examine and approve the project server before granting it access
  3. Confirm required environment variables are injected in the CI runtime
  4. Inspect a reviewed project .mcp.json definition into version control

Explanation

Correct Answer

A:

User-supplied identifiers are untrusted inputs, not authorization. The MCP server must scope data access to the authenticated principal to prevent cross-tenant disclosure.

Incorrect Answers

B:

Repository-supplied configurations can execute processes or contact services; that serves reviewing untrusted project MCP definitions, whereas the case requires enforcing tenant isolation at the MCP backend.

C:

Because an MCP configuration that expands environment references depends on that process's environment, the measure supports supplying MCP configuration variables in CI rather than the needed enforcing tenant isolation at the MCP backend.

D:

Project-scoped MCP configuration can be distributed with the repository; its benefits for sharing an MCP server configuration with a project do not resolve the need for enforcing tenant isolation at the MCP backend.

Question 161

A resumed session includes outdated authorization claims for an employee whose role changed. What should be checked again?

  1. Save a session identifier and resume its authorized conversation state
  2. Reauthenticate or reauthorize the user before privileged continuation
  3. Cross-check the persisted operation ID against the payment system before retrying
  4. Preserve decision timestamps, tool evidence and escalation events with the session

Explanation

Correct Answer

B:

A stored conversation is not a perpetual permission grant. Current identity and access checks must govern new tool actions after resumption.

Incorrect Answers

A:

Session continuity reconnects the agent to relevant history and progress; this may support resuming a saved agent session but cannot alone meet the requirement for revalidating authorization on session resume.

C:

A durable transcript does not guarantee whether the external effect committed; a method suited to reconciling external actions after session interruption does not meet the need for revalidating authorization on session resume.

D:

For preserving a session-level decision audit trail, auditable session state links outcomes to observations and approvals; however, this requirement concerns revalidating authorization on session resume.

Question 162

A developer asks the model to update code but no tests exist for the affected behavior. What step improves safety?

  1. Assemble reproduction inputs and compare observed environment differences first
  2. Build in the failing edge case to tests and implement the smallest correction
  3. Examine and checkpoint cohesive changes with their test evidence
  4. Add a focused regression test that reproduces the bug before or with the fix

Explanation

Correct Answer

D:

A targeted test captures the desired behavior and guards against future regressions. Success measured only by the generated explanation is not sufficient.

Incorrect Answers

A:

Without a reproducible failure or causal evidence, broad edits are speculative; that relates to establishing a reliable bug reproduction, not the necessary establishing a regression test for a repaired defect.

B:

Although treating review feedback as a concrete test turns a discovered omission into durable coverage, the choice addresses incorporating edge-case feedback into tests and not the requested establishing a regression test for a repaired defect.

C:

Small validated increments provide rollback boundaries and clearer blame history; the deciding condition is establishing a regression test for a repaired defect, whereas this action concerns checkpointing incremental implementation progress.

Question 163

A tool experiences a 429 response with a server-supplied delay hint. How should clients treat it?

  1. Tell apart a legitimate not-found result from a tool execution error
  2. Invoke again time-sensitive reads after recovery and replace stale failure state
  3. Provide a non-retryable validation error with the offending field identified
  4. Respect bounded retry-after guidance in the retry scheduler

Explanation

Correct Answer

D:

Rate-limit responses can indicate when another attempt is reasonable. Controlled backoff avoids synchronized retry storms and stays within service capacity.

Incorrect Answers

A:

A successful search with zero matches is not necessarily an infrastructure failure; in this scenario, separating no-match outcomes from failures is secondary to honoring rate-limit backoff information.

B:

When considering refreshing state after a transient service failure, cached failure artifacts are not live evidence; the requirement for honoring rate-limit backoff information remains unmet.

C:

When addressing distinguishing validation faults from transient errors, input-shape errors persist until arguments change; the present task instead calls for honoring rate-limit backoff information.

Question 164

An invoice pipeline needs standardized paid/overdue/disputed statuses and a safe way to represent an unknown field. Which TWO JSON schema choices help? Choose TWO.

  1. Represent absent invoice fields with an explicit nullable type and documented null convention
  2. Keep versioned the result contract and test consumers before deploying schema changes
  3. Insist on an actual authorized tool invocation and process its result through the dispatcher
  4. Constrain invoice statuses to the permitted paid, overdue and disputed enum values
  5. Find nonstandard stop or refusal outcomes before treating content as a schema result

Explanation

Correct Answers

A:

Alongside restricting status output to known categories, the workflow needs representing legitimate missing data in a schema: a deliberate null representation lets consumers distinguish unknown data from malformed output; asking the model to invent a value to satisfy a required non-null field harms correctness.

D:

Enumerated values constrain the output vocabulary and simplify downstream branching. Synonyms and typos in free text create unnecessary status-mapping failures.

Incorrect Answers

B:

Structured outputs create a stable interface that downstream systems depend upon; the action supports managing downstream compatibility of structured outputs but falls short of restricting status output to known categories in this case.

C:

A typed JSON response is not a network operation; this can serve distinguishing actual tool execution from JSON text, but leaves the requirement for restricting status output to known categories unmet.

E:

Schema-oriented APIs may have special refusal or termination paths; this concerns handling non-result terminations in structured workflows, leaving the requirement for restricting status output to known categories outstanding.

Question 165

One worker in a parallel research job times out, while two others succeed. What should the parent record?

  1. Verify the worker's claimed completion against authoritative tool status
  2. Track a partial workflow outcome and identify the unavailable branch explicitly
  3. Focus on propagating a sanitized error category and protected correlation reference
  4. Cross-check the external operation using an idempotency identifier before replay

Explanation

Correct Answer

B:

Partial success is not whole-task success. The coordinator can reuse completed outputs but must label the missing evidence and recovery need.

Incorrect Answers

A:

Because subagent prose should not override an explicit failed execution record, the measure supports verifying delegated completion against real error state rather than the needed preserving partial-failure status across worker aggregation.

C:

Error propagation should preserve recovery information without spreading secrets; supporting redacting sensitive details in propagated errors cannot replace the required preserving partial-failure status across worker aggregation.

D:

Distributed errors may arrive after a side effect has occurred; this contributes to preventing duplicate side effects in retrying workers without meeting the need for preserving partial-failure status across worker aggregation.

Question 166

A research report attributes a claim to a source URL that the search worker never opened. What should the pipeline demand?

  1. Limit the claim and request stronger corroboration before treating it as established
  2. Log retrieval time and a source version or approved snapshot reference
  3. Verify that every cited claim has an inspected supporting source
  4. Keep original-source identifiers and evidence type through every handoff

Explanation

Correct Answer

C:

A plausible citation string is not evidence of source support. Claim-to-passage checking guards against fabricated or misapplied references.

Incorrect Answers

A:

Weak or incomplete evidence warrants explicit uncertainty; a method suited to calibrating claims to evidence strength does not meet the need for checking citations against actually inspected evidence.

B:

For preserving time-sensitive research provenance, live pages can change or disappear; however, this requirement concerns checking citations against actually inspected evidence.

D:

A claim's evidentiary strength depends on where it originated; this may support retaining primary-source lineage across agents but cannot alone meet the requirement for checking citations against actually inspected evidence.

Question 167

One MCP connection offers sensitive admin operations and ordinary search to all agents. What architecture offers finer access boundaries?

  1. Specify each tool's selection criteria and record type in descriptions
  2. Expose separated tool sets or filtered capabilities according to worker identity
  3. Prefix or rename tools to identify their owning integration and purpose
  4. Block the unauthorized call and report a permission denial to orchestration

Explanation

Correct Answer

B:

Different trust levels deserve different callable surfaces. Role-based exposure can limit sensitive operations without losing routine integration features for lower-trust workers.

Incorrect Answers

A:

Tool routing quality depends on distinct metadata; the deciding condition is partitioning sensitive and ordinary MCP capabilities, whereas this action concerns disambiguating adjacent retrieval tools.

C:

Unique, descriptive identifiers help the agent and dispatcher target the intended backend; this addresses resolving tool name collisions, while the question is about partitioning sensitive and ordinary MCP capabilities.

D:

A capability boundary must be enforced at execution time; that relates to enforcing a denied capability at runtime, not the necessary partitioning sensitive and ordinary MCP capabilities.

Question 168

A developer wrote a file-pattern rule but used an extension not present anywhere in the repo. What QA should catch this?

  1. Test rule glob matches against real repository paths
  2. Choose a recursive path pattern matching Terraform files across modules
  3. Establish path-scoped rules under .claude/rules with file-pattern applicability
  4. Settle overlapping conditions and specify one unambiguous policy

Explanation

Correct Answer

A:

A simple match audit reveals dead rules that never load. Checking coverage is more reliable than assuming a syntactically valid pattern is useful.

Incorrect Answers

B:

Recursive globs represent a class of files regardless of module depth; covering nested infrastructure files consistently and verifying rule selectors actually match files require different controls in this situation.

C:

Path-conditioned repository guidance applies specialized instructions only to matching files; the unsatisfied requirement is verifying rule selectors actually match files, not applying code-style rules by path.

D:

Conflicting active instructions can make outcomes nondeterministic; in this scenario, preventing contradictory conditional instructions is secondary to verifying rule selectors actually match files.

Question 169

A pipeline validates field types but never checks whether order IDs exist. What additional layer is needed?

  1. Validate extracted references against authoritative business records
  2. Recompute totals and reject or review arithmetic inconsistencies
  3. Filter redundant outputs from extracted records using source identifiers and document context
  4. Check only that extracted order IDs appear as text in the document, without validating those IDs in business records

Explanation

Correct Answer

A:

Referential checks establish that identifiers and relationships are real. JSON schema validation alone cannot confirm existence in external systems.

Incorrect Answers

B:

Semantic checks compare extracted values to business invariants; that addresses checking arithmetic consistency in extracted records, whereas this case calls for checking referential integrity after extraction.

C:

Overlapping page captures can lead to double counting; this concerns removing duplicate extraction artifacts, leaving the requirement for checking referential integrity after extraction outstanding.

D:

A document can contain a syntactically plausible or invalid order number. Referential integrity requires looking up the ID in the authoritative order system, not just finding a text span.

Question 170

A team needs to retrieve batch results several months after processing without storing them itself. What retention issue must it address?

  1. Maintain each request's custom_id and associated error classification
  2. Persist results to controlled storage within the documented retrieval period
  3. Persist outcomes by input ID and create a retry queue from non-success statuses
  4. Select asynchronous batching and budget for deferred results

Explanation

Correct Answer

B:

Batch result URLs are not an archival store. The application must download and retain required records before provider retention windows elapse.

Incorrect Answers

A:

Stable item identifiers connect failures to inputs and retry history; supporting linking each batch error to its source record cannot replace the required preserving batch output beyond service retention.

C:

Structured reconciliation lets teams preserve successful results and focus intervention on failures; that serves constructing targeted post-batch retries, whereas the case requires preserving batch output beyond service retention.

D:

Because batch processing can reduce unit cost at the expense of immediate completion, the measure supports trading response latency for bulk-processing cost rather than the needed preserving batch output beyond service retention.

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