Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 36

Question 350

The investigative tools have resolved the question and the service must return a concise conclusion. Which production safeguard is most relevant?

  1. Finish with a user-facing answer after the last action
  2. Escalate after repeated inconclusive tool results
  3. Observe a tool result before choosing the next action
  4. Use a light direct API integration

Explanation

Correct Answer

A:

The model synthesizes a final response once no additional client tool request is needed; notably, a raw tool result is not automatically a contextualized customer answer.

Incorrect Answers

B:

Bounded fallback returns uncertainty and a trace for human follow-up rather than fabricating certainty; in contrast, the model synthesizes a final response once no additional client tool request is needed, while extra iterations are not justified once the cost and confidence guardrails trigger.

C:

Alternating action and observation permits evidence-driven choices rather than a fully predetermined plan; this addresses a separate concern, not the fact that the model synthesizes a final response once no additional client tool request is needed; every loop must check for stop conditions and tool failures.

D:

A direct client minimizes abstraction and makes the request path easy to inspect; here, the controlling capability is that the model synthesizes a final response once no additional client tool request is needed, but it will need explicit orchestration if tool calls and branching emerge later.

Question 351

Users provide scripts for analysis and an automated workflow may execute them during evaluation. Which approach best meets this requirement?

  1. Monitor and investigate unusual tool activity
  2. Keep a human review gate for high-impact operations
  3. Use a policy-based refusal and safe redirect
  4. Sandbox uncertain code execution

Explanation

Correct Answer

D:

Runtime isolation limits privileges, file access, and network effects of untrusted programs; the implementation must still recognize that static text review cannot guarantee the code will behave safely.

Incorrect Answers

A:

Telemetry and alerts identify suspicious action patterns and allow prompt, tool, or policy remediation; this cannot substitute for the behavior that runtime isolation limits privileges, file access, and network effects of untrusted programs, because monitoring should complement, not replace, preventive execution controls.

B:

Mandatory human approval separates recommendations from irreversible execution; in contrast, runtime isolation limits privileges, file access, and network effects of untrusted programs, while model confidence scores cannot replace the required authority of a reviewer.

C:

A policy layer handles the request appropriately and communicates a safe boundary; this addresses a separate concern, not the fact that runtime isolation limits privileges, file access, and network effects of untrusted programs; policy adherence should be evaluated; it is not guaranteed by a single warning prompt.

Question 352

Claude frequently confuses two inventory functions whose descriptions are vague and their parameters overlap. Which approach best meets this requirement?

  1. Provide precise tool descriptions and argument types
  2. Represent allowed values as enums where appropriate
  3. Read the returned tool_use ID and name
  4. Mark truly required arguments in the schema

Explanation

Correct Answer

A:

Clear names, schema fields, and descriptions help the model choose and populate the intended operation; notably, tool schemas still need server-side validation of all inputs.

Incorrect Answers

B:

Constrained value sets prevent unsupported strings from reaching backend business logic; in contrast, clear names, schema fields, and descriptions help the model choose and populate the intended operation, while valid enum membership alone cannot authorize an illegal state transition.

C:

Each client tool request includes metadata that ties subsequent execution and results to that call; this addresses a separate concern, not the fact that clear names, schema fields, and descriptions help the model choose and populate the intended operation; guessing an operation from nearby text can dispatch the wrong function.

D:

The schema identifies mandatory fields so calls missing them can be rejected before dispatch; here, the controlling capability is that clear names, schema fields, and descriptions help the model choose and populate the intended operation, but optional property defaults should not hide missing business-critical identifiers.

Question 353

A submitted batch may return individual results in an order different from its source file. What would best address the stated failure mode?

  1. Use the real-time Messages API for chat
  2. Handle partial failures inside a batch
  3. Correct a malformed request before resubmission
  4. Correlate batch results by custom identifiers

Explanation

Correct Answer

D:

Stable request IDs let the application map each result to the correct original record; the important constraint is that relying on result ordering can mix customer records or output destinations.

Incorrect Answers

A:

Direct Messages calls match interactive latency and incremental display requirements; by comparison, the intended solution relies on this: stable request IDs let the application map each result to the correct original record; batch jobs complete later and are inappropriate for a tightly coupled request-response interface.

B:

Per-item result handling records and retries appropriate failures without duplicating successful work; the needed capability is different—stable request IDs let the application map each result to the correct original record—and a batch-level completion event does not imply every included request succeeded.

C:

Client-side schema validation fixes the payload and prevents fruitless retries, whereas the requirement calls for another behavior: stable request IDs let the application map each result to the correct original record; a network backoff cannot make an invalid request body well formed.

Question 354

A payment assistant returns a valid numeric amount but it exceeds the customer's authorization limit. What is the best fit for the described situation?

  1. Apply business rules after schema checks
  2. Require approval for write statements
  3. Parse generated JSON with a real parser
  4. Handle invalid responses with bounded repair

Explanation

Correct Answer

A:

Independent business validation enforces domain-specific ranges and account scope; the implementation must still recognize that pure JSON Schema cannot determine whether a payment is authorized.

Incorrect Answers

B:

A controlled executor enforces authorized write permissions and human approval where required; the missing capability is more specific: independent business validation enforces domain-specific ranges and account scope; a syntax-valid UPDATE statement is not proof of business permission.

C:

Standard JSON parsing catches syntactic errors instead of relying on visual plausibility; by comparison, the intended solution relies on this: independent business validation enforces domain-specific ranges and account scope; successfully parsed JSON can still violate expected business types.

D:

A controlled repair or regeneration step may recover a valid shape without pretending the first response was correct; the needed capability is different—independent business validation enforces domain-specific ranges and account scope—and endless retries can increase cost without resolving a faulty contract.

Question 355

One developer wants a preferred explanatory style across different projects without modifying each repository. Which option best fits the described workflow?

  1. Use local settings for workstation-only behavior
  2. Maintain repository CLAUDE.md instructions
  3. Run a configuration-diff preflight
  4. Use user-level guidance for personal defaults

Explanation

Correct Answer

D:

Personal configuration scopes recurring individual preferences independently of shared project rules; notably, private defaults should not override necessary repository safety requirements.

Incorrect Answers

A:

Local configuration can hold machine-specific client preferences outside shared version control, whereas the requirement calls for another behavior: personal configuration scopes recurring individual preferences independently of shared project rules; committing private machine policy could surprise teammates.

B:

Project memory files make scoped guidance available in Claude Code working contexts; however, the critical mechanism here is that personal configuration scopes recurring individual preferences independently of shared project rules, and they should not contain credentials or private production secrets.

C:

Automated checks compare effective values against the approved deployment baseline; the requirement instead turns on this behavior: personal configuration scopes recurring individual preferences independently of shared project rules; assuming staging and production are identical is not a verification method.

Question 356

A developer wants persistent policy rules that ordinary user messages cannot rewrite directly. Which of these alternatives serves the stated goal?

  1. Put customer questions in user messages
  2. Use application-controlled system instructions
  3. Treat attacker content as quoted data only
  4. Use tool schemas for action interfaces

Explanation

Correct Answer

B:

The system-level field communicates application guidance separately from task content; a crucial qualification is that backend authorization still must be enforced in code and tools.

Incorrect Answers

A:

The user role carries the requested task without altering the registered tool definitions; here, the controlling capability is that the system-level field communicates application guidance separately from task content, but making every customer turn a system rule would collapse trust separation.

C:

The application preserves the page as untrusted input and excludes it from control instructions; by comparison, the intended solution relies on this: the system-level field communicates application guidance separately from task content; a malicious instruction can still be persuasive, so execution permissions remain bounded.

D:

Tool definitions and server validation specify callable operations more reliably than prose parsing; the needed capability is different—the system-level field communicates application guidance separately from task content—and a natural-language command is not an adequate machine-checked action contract.

Question 357

Before enabling a customer-supplied script-analysis feature, a team must constrain what untrusted code can access and test whether retrieved documents can trick the assistant into unsafe actions. Which TWO protections tackle execution containment and adversarial evaluation? Choose TWO.

  1. Sandbox uncertain code execution
  2. Canary a risky policy change
  3. Run adversarial prompt-injection evaluations
  4. Use constrained tool schemas and allowlists
  5. Inspect outputs before taking business actions

Explanation

Correct Answers

A:

Runtime isolation limits privileges, file access, and network effects of untrusted programs; in production, static text review cannot guarantee the code will behave safely.

C:

An abuse test suite probes trust boundaries and confirms tools remain protected, although passing a few normal demonstrations does not measure adversarial resilience.

Incorrect Answers

B:

A limited rollout measures refusal, error, and unauthorized action attempts before broad exposure; the needed capability is different—runtime isolation limits privileges, file access, and network effects of untrusted programs—and immediate all-user changes make regressions harder to contain.

D:

Narrow tool capabilities reduce what even a manipulated model could attempt; however, the critical mechanism here is that runtime isolation limits privileges, file access, and network effects of untrusted programs, and a policy paragraph in system instructions is not a technical allowlist.

E:

Output checks validate schema, scope, and safety before any backend mutation; the requirement instead turns on this behavior: runtime isolation limits privileges, file access, and network effects of untrusted programs; post-generation checks cannot repair unauthorized access already performed upstream.

Question 358

An agent might call an order-placement endpoint twice because the first confirmation was lost. Which capability is necessary for this outcome?

  1. Attach an idempotency token to a business write
  2. Reject unexpected fields in business actions
  3. Retry only failures considered transient
  4. Use capped exponential backoff with jitter

Explanation

Correct Answer

A:

Stable operation identity lets the transactional endpoint suppress duplicate side effects; a crucial qualification is that randomizing the token on every retry defeats the protection.

Incorrect Answers

B:

Application validation controls the allowed property set before any database mutation; by comparison, the intended solution relies on this: stable operation identity lets the transactional endpoint suppress duplicate side effects; the model may produce syntactically valid extra keys that are unsafe to execute.

C:

A retry policy classifies failure causes and retries only the recoverable subset; the needed capability is different—stable operation identity lets the transactional endpoint suppress duplicate side effects—and repeating all failures indiscriminately can amplify outages and costs.

D:

Randomized increasing delays spread load and respect retry guidance where supplied, whereas the requirement calls for another behavior: stable operation identity lets the transactional endpoint suppress duplicate side effects; unbounded retries create unpredictable latency and can hide an enduring fault.

Question 359

Auditors must verify whether a generated conclusion follows the current company policy text. What would best address the stated failure mode?

  1. Authorize each requested resource operation
  2. Refresh an index when content changes
  3. Preserve source references alongside snippets
  4. Apply authorization before retrieval

Explanation

Correct Answer

C:

Provenance metadata allows citations and later source checks for each retrieved excerpt; in production, a plausible paragraph with no reference cannot be independently traced.

Incorrect Answers

A:

Resource-level policy checks verify that the caller can execute that exact action on that object; here, the controlling capability is that provenance metadata allows citations and later source checks for each retrieved excerpt, but authentication alone does not confer access to all records.

B:

A freshness pipeline updates search representations and source versions when documents change; the missing capability is more specific: provenance metadata allows citations and later source checks for each retrieved excerpt; more prompt instructions cannot correct an out-of-date underlying index.

D:

The retrieval layer filters sources by verified caller access before any text reaches Claude; the needed capability is different—provenance metadata allows citations and later source checks for each retrieved excerpt—and filtering only the final answer cannot undo earlier disclosure to the model.

Question 360

The API responds with a credential-related authorization failure for a deployed worker. What is the most direct implementation choice?

  1. Correct a malformed request before resubmission
  2. Refresh authentication on unauthorized calls
  3. Treat repeated server errors as a dependency outage
  4. Register a client tool with input_schema

Explanation

Correct Answer

B:

The operator checks key validity, workspace scope, and secret injection before retrying; notably, sending the same invalid secret repeatedly will not solve the auth problem.

Incorrect Answers

A:

Client-side schema validation fixes the payload and prevents fruitless retries; the requirement instead turns on this behavior: the operator checks key validity, workspace scope, and secret injection before retrying; a network backoff cannot make an invalid request body well formed.

C:

A circuit breaker and fallback degrade gracefully while limiting repeated failed traffic; in contrast, the operator checks key validity, workspace scope, and secret injection before retrying, while a longer prompt or higher max_tokens does not restore an unavailable upstream server.

D:

Tool definitions describe names, parameters, and types so Claude can propose structured calls; this addresses a separate concern, not the fact that the operator checks key validity, workspace scope, and secret injection before retrying; defining a schema does not grant access to an external system automatically.

Claude Certified Developer - Foundations (CCDV-F) Practice Tests