McAfee Secure

Exam Code: CCA-F

Exam Name: Claude Certified Architect - Foundations

Certification Provider: Anthropic

Questions & Answers

Anthropic CCA-F Questions & Answers

Study with Up-To-Date REAL Exam Questions and Answers from the ACTUAL Test

Questions & Answers

134 Questions & Answers with Testing Engine

"Claude Certified Architect - Foundations Exam", also known as CCA-F exam, is a Anthropic certification exam.

Pass your tests with the always up-to-date CCA-F Exam Engine. Your CCA-F training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Anthropic CCA-F Practice Test 16

Question 150

A client config forces tool_choice to a function name that is not in the registered tool definitions. What should the host do before sending the request?

  1. Disable the schema checks to allow arbitrary strings.
  2. Substitute the most privileged tool to ensure a call occurs.
  3. Ask Claude to guess a similarly spelled tool name at runtime.
  4. Validate the forced name against declared tools and resolve or report the configuration mismatch.

Explanation

Correct Answer

D:

An unregistered target cannot be invoked reliably.

Incorrect Answers

A:

Malformed tool declarations cannot create missing execution logic.

B:

Capability escalation is not a safe fallback.

C:

Guessing may select the wrong operation.

Question 151

Two subagents have overlapping tool descriptions for read_policy and change_policy, and the verifier selects the mutating one. What is the best correction?

  1. Give the verifier approval rights to compensate for selection errors.
  2. Ask the verifier to make a harmless write before every read.
  3. Rename both tools policy and rely on context clues.
  4. Differentiate read and write contracts clearly and remove the mutating tool from the verifier allowlist.

Explanation

Correct Answer

D:

Distinct labels plus permissions prevent side effects during validation.

Incorrect Answers

A:

More authority would worsen the consequence of a mistake.

B:

Extra state changes are not needed for verification.

C:

Generic naming increases ambiguity.

Question 152

A summarization agent has all facts in its message and needs no external action. Its controller forces tool_choice any, causing a useless call. What change is justified?

  1. Allow a normal text response for tool-free tasks instead of requiring an unnecessary invocation.
  2. Remove all source facts from the model context so it must search again.
  3. Force a file deletion to satisfy the API requirement for a call.
  4. Repeat a read of the same unchanged source every turn.

Explanation

Correct Answer

A:

Tool-use constraints should reflect actual data or side-effect dependencies.

Incorrect Answers

B:

Discarding verified evidence creates artificial work.

C:

An unrelated mutation is hazardous.

D:

Redundant fetches waste resources without improving the summary.

Question 153

A documentation platform exposes thousands of known documents. A client needs discovery and reading, not an action to mutate each listing. Select TWO decisions that make the document catalog safe and discoverable through MCP.

  1. Expose every document as a refund or delete tool.
  2. Return only an unstructured blob with no identifiers.
  3. Require models to guess URLs rather than discover available resources.
  4. Expose discoverable read-only MCP resources with stable identifiers for document content and listings.
  5. Authorize resource discovery and reads per authenticated client, so the catalog does not expose another tenant’s documents.

Explanation

Correct Answers

D:

Resources support contextual access without pretending every document is an action.

E:

A stable resource identifier makes documents retrievable, but it must still be filtered by the calling principal’s permissions. The server must not treat the catalog as a global public list of all tenants’ files.

Incorrect Answers

A:

Mutating tool semantics do not fit static documents.

B:

Clients cannot reliably retrieve specific references.

C:

Guesswork weakens provenance and access control.

Question 154

An MCP server supports reading a catalog and issuing refunds. Which operation should be an authorized tool instead of only a resource?

  1. Publish refunds as static reference data, with no action semantics.
  2. Combine catalog listing with automatic refund on every read.
  3. Allow reading catalog descriptions to imply refund authorization.
  4. Issue_refund as a typed, permission-checked tool because it changes payment state.

Explanation

Correct Answer

D:

State-changing operations require explicit arguments and execution controls.

Incorrect Answers

A:

A document resource cannot commit a financial transaction.

B:

A read request should never trigger payment side effects.

C:

Access to information does not grant business action rights.

Question 155

An engineering group needs issue-tracker integration and finds a maintained MCP server matching its read/write requirements. What is the sensible starting decision?

  1. Connect the available server with unrestricted tokens and no review.
  2. Evaluate the maintained server's security, schemas, permissions and reliability before building a custom connector.
  3. Build a new server immediately without reviewing the existing one.
  4. Disable audit logging because the server is maintained by another party.

Explanation

Correct Answer

B:

Reuse can save work when the existing implementation meets the operational contract.

Incorrect Answers

A:

Convenience does not justify broad secrets and unchecked tools.

C:

Unnecessary bespoke infrastructure adds maintenance cost.

D:

External maintenance does not remove internal accountability.

Question 156

A company must expose a private shipment API for which no suitable maintained MCP integration exists. What should the architect build?

  1. Write a tool description that claims access but executes no API call.
  2. A narrowly scoped custom MCP server with typed operations, authentication, authorization and predictable errors.
  3. Give the model the shipment database credentials and ask it to connect directly.
  4. Proxy arbitrary URLs with a shared administrator token.

Explanation

Correct Answer

B:

A purpose-built adapter can expose approved capabilities without granting raw API access.

Incorrect Answers

A:

Documentation cannot replace an implementation.

C:

Secrets and unrestricted database access bypass the tool boundary.

D:

Unbounded proxying creates serious access and network risks.

Question 157

An MCP server receives a secret key through tool arguments and forwards it into agent-visible logs. Select TWO security repairs for the exposed MCP credential flow.

  1. Manage credentials server-side and expose only least-privilege, task-specific operations to the model.
  2. Share one master token across every user and tenant.
  3. Return the secret in a debug field after every call.
  4. Put the credential in the system prompt so the agent never forgets it.
  5. Remove key values from model-visible logs and rotate any credential already exposed by the faulty tool path.

Explanation

Correct Answers

A:

Server-held secrets reduce prompt and logging exposure.

E:

Moving future secrets into server-side management stops new leakage. Redaction and rotation address the separate exposure that already happened in earlier logs.

Incorrect Answers

B:

Shared privileges defeat meaningful isolation.

C:

Unnecessary disclosure expands the incident surface.

D:

Prompts are not secure secret stores.

Question 158

A shared MCP customer-data server is callable by agents for many customers. It trusts the account_id argument and exposes any tenant's records. Which safeguard is missing?

  1. Grant all users reader access to all accounts for consistent behavior.
  2. Hide tenant IDs from the front-end but continue accepting arbitrary values.
  3. Enforce authenticated principal and tenant-scoped authorization on every server-side resource and tool request.
  4. Ask the model to promise to use only its own tenant's data.

Explanation

Correct Answer

C:

Caller-supplied record IDs are not proof of permission.

Incorrect Answers

A:

Consistency cannot justify cross-tenant disclosure.

B:

Concealment does not block unauthorized reads.

D:

Prompt instructions cannot enforce multitenancy isolation.

Question 159

A model needs to read an existing reference document and also submit a new reimbursement. How should the MCP interface distinguish these?

  1. Make document reading automatically submit reimbursement as a convenience.
  2. Expose the stable document as a resource and the reimbursement submission as a separately authorized action tool.
  3. Treat both operations as untyped text outputs with no identity tracking.
  4. Present the reimbursement endpoint as a static resource with no executable contract.

Explanation

Correct Answer

B:

The data/read boundary differs from a state-mutating business operation.

Incorrect Answers

A:

Read access must not trigger payment mutation.

C:

Typed input and side-effect contracts are necessary for accountability.

D:

A resource alone cannot perform the requested state change.

Question 160

A Claude Code team wants all contributors to use the same approved MCP test-server configuration from the repository. What deployment scope fits?

  1. Use a personal machine-only configuration and assume the team inherits it.
  2. Place every teammate's API token directly in the project config.
  3. Use version-controlled project-scoped MCP configuration while keeping individual credentials outside committed files.
  4. Copy the server binary into the prompt whenever a session starts.

Explanation

Correct Answer

C:

Shared configuration improves reproducibility without exposing personal secrets.

Incorrect Answers

A:

User-local files do not automatically appear on other machines.

B:

Committing credentials would expose privileged access.

D:

Prompts are not a reliable deployment mechanism.

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