Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 41

Question 400

A healthcare support assistant only needs delivery status, while its connector returns sensitive identity fields and its traces may include private headers. Which TWO data-handling measures reduce exposure before inference and during logging? Choose TWO.

  1. Minimize personal fields before inference
  2. Apply retention and deletion rules to transcripts
  3. Authorize access to stored conversation history
  4. Treat a retrieved webpage as untrusted data
  5. Redact secrets from observability logs

Explanation

Correct Answers

A:

Data minimization removes unrelated personal information prior to sending content to the model; a request to disregard a field still exposes that field if it is transmitted.

E:

Redaction removes sensitive fields before events reach metrics and log sinks; notably, encrypting logs does not justify logging unnecessary secrets indefinitely.

Incorrect Answers

B:

Storage controls govern how long records, logs, and derived artifacts remain available, whereas the requirement calls for another behavior: data minimization removes unrelated personal information prior to sending content to the model; a model's context window does not define legal data retention requirements.

C:

Access-control checks at the storage layer prevent cross-account conversation disclosure; however, the critical mechanism here is that data minimization removes unrelated personal information prior to sending content to the model, and a hidden session identifier is not enough to authorize retrieval.

D:

External content is evidence, not an instruction source allowed to override the application's policies; the requirement instead turns on this behavior: data minimization removes unrelated personal information prior to sending content to the model; prompt labeling helps but tool-side access restrictions must remain enforced.

Question 401

An audited extraction workflow must not silently change behavior when a friendly model alias moves. Which measure directly addresses the problem?

  1. Keep rollback artifacts and data migration notes
  2. Set the requested API version header where required
  3. Document breaking API changes for callers
  4. Pin a tested exact model version when necessary

Explanation

Correct Answer

D:

Version pinning keeps the model selection stable until the team deliberately upgrades and retests; the implementation must still recognize that the pinned model may later require migration if the provider retires it.

Incorrect Answers

A:

Release planning preserves prior binaries and any required data conversion strategy; this cannot substitute for the behavior that version pinning keeps the model selection stable until the team deliberately upgrades and retests, because reverting only source files may not undo incompatible database changes.

B:

Version negotiation constrains accepted fields and response behavior for an endpoint; in contrast, version pinning keeps the model selection stable until the team deliberately upgrades and retests, while omitting required version information or assuming an arbitrary revision can break compatibility.

C:

A versioned change notice and migration path help clients adopt the new contract safely; this addresses a separate concern, not the fact that version pinning keeps the model selection stable until the team deliberately upgrades and retests; a silent rollout may turn valid old requests into failures.

Question 402

A queue contains thousands of pending inference jobs but the organization has finite rate limits. Which design would meet this acceptance criterion?

  1. Use async I/O for many concurrent inference requests
  2. Bound concurrency with a semaphore
  3. Use capped exponential backoff with jitter
  4. Use a work queue for long offline jobs

Explanation

Correct Answer

B:

A concurrency limit caps simultaneous API work and prevents uncontrolled bursts, subject to this requirement: launching every coroutine at once can trigger rate limiting and memory pressure.

Incorrect Answers

A:

Asynchronous clients yield during I/O and resume when responses arrive; in contrast, a concurrency limit caps simultaneous API work and prevents uncontrolled bursts, while an event loop does not make a single model inference computationally faster.

C:

Randomized increasing delays spread load and respect retry guidance where supplied; here, the controlling capability is that a concurrency limit caps simultaneous API work and prevents uncontrolled bursts, but unbounded retries create unpredictable latency and can hide an enduring fault.

D:

A persistent queue decouples submission from workers and supports retry state; the missing capability is more specific: a concurrency limit caps simultaneous API work and prevents uncontrolled bursts; keeping a web request open indefinitely is fragile under client disconnects.

Question 403

Public web research and private customer data require different tool permissions and data exposure rules. What is the best fit for the described situation?

  1. Ask a clarifying question when critical input is missing
  2. Split by security boundary
  3. Avoid splitting when coordination overhead dominates
  4. Split by independent artifact

Explanation

Correct Answer

B:

Isolation by trust zone prevents a public-facing worker from gaining private data access; worker separation without separate credentials or tool allowlists is superficial.

Incorrect Answers

A:

Requesting an authorized scope prevents guesses from becoming unintended side effects; by comparison, the intended solution relies on this: isolation by trust zone prevents a public-facing worker from gaining private data access; an internal plan cannot legitimately invent the missing tenancy information.

C:

Keeping a bounded task local avoids orchestration latency and reconciliation errors, whereas the requirement calls for another behavior: isolation by trust zone prevents a public-facing worker from gaining private data access; subagents should be justified by distinct work or isolation needs, not mere availability.

D:

Artifact-level workers can process distinct diffs in parallel and return scoped results; however, the critical mechanism here is that isolation by trust zone prevents a public-facing worker from gaining private data access, and shared global context might cause one change set to bias another.

Question 404

A user asks for current inventory quantities without providing any source or tool. Which mechanism should the team prioritize?

  1. Treat output as generated prediction, not a database read
  2. Use a computation tool for exact arithmetic
  3. Estimate token usage with supported token counting
  4. Use retrieval for current factual evidence

Explanation

Correct Answer

A:

The language model generates plausible text rather than querying a company's live inventory by itself, subject to this requirement: it needs authorized retrieval to ground fast-changing private facts.

Incorrect Answers

B:

A trusted arithmetic implementation calculates the numeric result instead of relying on next-token prediction; here, the controlling capability is that the language model generates plausible text rather than querying a company's live inventory by itself, but the model can explain the calculation but should not be the authoritative calculator.

C:

Token counting estimates model tokens more accurately than character length or word count; the missing capability is more specific: the language model generates plausible text rather than querying a company's live inventory by itself; different formats and models may tokenize the same text differently.

D:

A retrieval step supplies updated source text before answer generation; by comparison, the intended solution relies on this: the language model generates plausible text rather than querying a company's live inventory by itself; retrieved content still needs authenticity and relevance checks.

Question 405

A structured response is incomplete and the product needs a safe retry before treating it as data. Which solution matches the actual constraint?

  1. Cross-check numerical fields with deterministic logic
  2. Validate values against a schema
  3. Handle invalid responses with bounded repair
  4. Parse generated JSON with a real parser

Explanation

Correct Answer

C:

A controlled repair or regeneration step may recover a valid shape without pretending the first response was correct; a crucial qualification is that endless retries can increase cost without resolving a faulty contract.

Incorrect Answers

A:

Code verifies arithmetic consistency after the model proposes the values; this addresses a separate concern, not the fact that a controlled repair or regeneration step may recover a valid shape without pretending the first response was correct; model confidence does not establish accounting correctness.

B:

Schema validation rejects type and enumeration mismatches before downstream use; here, the controlling capability is that a controlled repair or regeneration step may recover a valid shape without pretending the first response was correct, but the model being instructed to use JSON is not a runtime guarantee.

D:

Standard JSON parsing catches syntactic errors instead of relying on visual plausibility; by comparison, the intended solution relies on this: a controlled repair or regeneration step may recover a valid shape without pretending the first response was correct; successfully parsed JSON can still violate expected business types.

Question 406

A developer considers adding a service token beside project instructions to make tool access convenient. What is the correct engineering decision?

  1. Place narrow guidance close to the relevant code
  2. Keep secrets outside coding-agent memory files
  3. Initialize repository guidance deliberately
  4. Use a reusable Skill for repeatable know-how

Explanation

Correct Answer

B:

Credentials remain in protected stores or environment controls rather than committed textual guidance; in production, a shared file can be copied, logged, and inherited by unintended readers.

Incorrect Answers

A:

Scoped project guidance can communicate local conventions without overloading every task with unrelated instructions; the missing capability is more specific: credentials remain in protected stores or environment controls rather than committed textual guidance; conflicting guidance across scopes should be reviewed for effective precedence.

C:

An initialization workflow can help establish and review useful CLAUDE.md guidance for the repository; the needed capability is different—credentials remain in protected stores or environment controls rather than committed textual guidance—and generated notes must be checked before being treated as authoritative development policy.

D:

A Skill packages reusable task guidance so the coding agent can apply a consistent procedure, whereas the requirement calls for another behavior: credentials remain in protected stores or environment controls rather than committed textual guidance; guidance does not grant access to a tool or enforce backend authorization.

Question 407

A generated description will be inserted into a browser HTML view without executing user-supplied markup. Which choice should be used here?

  1. Encode JSON using a proper serializer
  2. Use server-sent events for streamed responses
  3. Escape output for its destination context
  4. Preserve UTF-8 content safely

Explanation

Correct Answer

C:

Context-aware escaping prevents text from becoming active script or HTML—with this limitation: JSON escaping is not interchangeable with HTML or SQL escaping.

Incorrect Answers

A:

A JSON encoder produces correctly escaped structured payloads, whereas the requirement calls for another behavior: context-aware escaping prevents text from becoming active script or HTML; manually concatenated JSON strings are vulnerable to malformed syntax and injection bugs.

B:

SSE carries incremental events from the service to the client as generation proceeds; however, the critical mechanism here is that context-aware escaping prevents text from becoming active script or HTML, and it is not a general-purpose bidirectional client messaging transport.

D:

Proper encoding preserves characters consistently between client, API, and storage; this cannot substitute for the behavior that context-aware escaping prevents text from becoming active script or HTML, because re-encoding bytes as the wrong character set can corrupt meaning.

Question 408

A maintenance assistant is prohibited from receiving deployment-write tools, and a rare permitted infrastructure action must still require a named person's explicit approval. Which TWO safeguards independently meet those conditions? Choose TWO.

  1. Use tool allowlists as a baseline defense
  2. Use a post-tool hook for outcome auditing
  3. Use a mandatory human approval mechanism
  4. Use a pre-tool hook for prohibited actions

Explanation

Correct Answers

A:

Tool availability and credential permissions restrict the reachable attack surface by design, subject to this requirement: hooks are additional safeguards, not replacements for minimizing available power.

C:

An enforced approval boundary stops execution until confirmation is provided; a crucial qualification is that a model's promise to ask first is not itself a technical approval gate.

Incorrect Answers

B:

A lifecycle hook can capture sanitized execution results and support monitoring or follow-up checks; here, the controlling capability is that tool availability and credential permissions restrict the reachable attack surface by design, but a post-action hook cannot retroactively prevent a harmful command that already ran.

D:

A pre-execution hook can inspect and block the operation before it reaches the shell; by comparison, the intended solution relies on this: tool availability and credential permissions restrict the reachable attack surface by design; the hook runs outside the model's judgment, but its own policy must be correctly implemented.

Question 409

A deployment may introduce parsing errors and the team must revert without data loss. Which action most directly closes the gap?

  1. Write a rollback plan
  2. Pin reproducible dependencies
  3. Require pull-request review
  4. Write a design review before implementation

Explanation

Correct Answer

A:

A rollback procedure identifies safe prior artifacts, migration constraints, and trigger thresholds; in production, reverting an application binary may be insufficient after incompatible schema changes.

Incorrect Answers

B:

Lockfiles and controlled versions reduce unplanned changes between test and production; the needed capability is different—a rollback procedure identifies safe prior artifacts, migration constraints, and trigger thresholds—and automatically taking every new major dependency version can break stable interfaces.

C:

Peer review examines diff, tests, and ownership before a controlled merge, whereas the requirement calls for another behavior: a rollback procedure identifies safe prior artifacts, migration constraints, and trigger thresholds; direct unreviewed edits reduce accountability and increase regression risk.

D:

Architecture review assesses interfaces, failure modes, and control ownership before code changes; however, the critical mechanism here is that a rollback procedure identifies safe prior artifacts, migration constraints, and trigger thresholds, and implementing immediately can lock in unsafe assumptions and costly redesign.

Question 410

An assistant reads a web page that includes instructions telling it to send secret credentials elsewhere. Which response is most defensible under these conditions?

  1. Monitor retrieval coverage and citation fidelity
  2. Keep user instructions subordinate to security policy
  3. Use tool results as data to be interpreted
  4. Treat retrieved pages as untrusted evidence

Explanation

Correct Answer

D:

External page text should be handled as data, not as new operational authority; the important constraint is that a page author cannot legitimately override the application's system policy.

Incorrect Answers

A:

Measuring evidence relevance and attribution identifies grounding defects; by comparison, the intended solution relies on this: external page text should be handled as data, not as new operational authority; extra context tokens cannot guarantee faithful citation of the retrieved material.

B:

The application enforces identity and authorization checks despite the request's wording; the needed capability is different—external page text should be handled as data, not as new operational authority—and requests from end users do not grant new backend privileges.

C:

Result content is evidence from an external subsystem and must not become a system instruction, whereas the requirement calls for another behavior: external page text should be handled as data, not as new operational authority; even trusted services can carry attacker-controlled record values.

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