Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 29

Question 280

A batch processing report displays only the number of completed records. Which KPI helps compare efficiency across versions?

  1. Introduce model-appropriate output limits and request budgets at the server.
  2. Distinguish ordinary input, cache writes, cache reads, and output tokens in reporting.
  3. Retain the policy in a stable prompt prefix and enable prompt caching.
  4. Track total spend and tokens per accepted record alongside quality.

Explanation

Correct Answer

D:

A version may complete more records at much higher token cost or error rate; accepted-output economics reveal the real tradeoff.

Incorrect Answers

A:

A server-side maximum constrains possible generated tokens independently of user requests, although the application must handle truncation; when the goal is normalized batch inference cost, optimizing spend guardrail per request misses that operational boundary.

B:

Cache operations can have different prices from uncached input, so a single input count misstates actual spend; this targets cache-aware chargeback, while the system must demonstrate normalized batch inference cost.

C:

Repeated prefixes can be read from cache at reduced input cost; dynamic requests should follow the cached material; the scenario's controlling requirement is normalized batch inference cost, not the related repeated policy input savings concern.

Question 281

A production model has frequent output retries caused by malformed data. What cost improvement should be considered?

  1. For this workflow, cache prefixes include earlier tool definitions, so changed tools can invalidate later entries.
  2. Log total spend and tokens per accepted record alongside quality.
  3. Use supported structured output contracts and validation to reduce wasteful generations.
  4. Handle cached prompts as time-limited entries and measure whether request intervals permit reuse.

Explanation

Correct Answer

C:

Preventing recoverable format failures can save retries and downstream processing, while semantic validation remains necessary.

Incorrect Answers

A:

Stable prompt text alone is insufficient when the overall preceding request structure changes; this targets tools as cache prefix dependency, while the system must demonstrate cost of output-format retries.

B:

A version may complete more records at much higher token cost or error rate; accepted-output economics reveal the real tradeoff; it improves normalized batch inference cost only, leaving cost of output-format retries unproven.

D:

Cache entries are not permanent storage; actual retention and reuse windows affect hit rates and savings even when prompt bytes and tool definitions stay unchanged; this may help cache lifetime and workload reuse interval but cannot by itself deliver cost of output-format retries.

Question 282

A support assistant stores a compressed memory that says a customer still prefers email, but the customer now explicitly requests phone contact. What context-management rule should the application apply?

  1. Append the new phone request but keep the contradictory old preference as equally active.
  2. Delete every other customer memory entry so no historical context can influence future requests.
  3. Treat the latest explicit customer preference as current and update the durable summary after verification.
  4. Always obey the old memory summary because it was included earlier in the prompt.

Explanation

Correct Answer

C:

Compressed history is a memory aid, not an authority that can override newer user intent. The application should preserve the updated preference and avoid reintroducing superseded information from old summaries.

Incorrect Answers

A:

Conflicting state makes later turns ambiguous and can cause unpredictable contact decisions. The memory record should distinguish current preference from historical change.

B:

A single changed preference does not justify discarding unrelated verified constraints. Updating the affected field preserves useful durable context without retaining the stale value as active.

D:

Earlier cached or summarized preferences may no longer be accurate. They should not outrank a clear current request from the authorized customer.

Question 283

A research assistant must inspect a hundred articles but answer only one narrow question. What should enter the main model context?

  1. Retrieve and include the most relevant excerpts with source identifiers.
  2. Save the address in protected order-scoped state with appropriate expiry.
  3. Put the complete text of every authorized article into the prompt before ranking them.
  4. Retrieve just-in-time code excerpts and expand only when evidence requires it.

Explanation

Correct Answer

A:

Targeted evidence reduces input tokens and distractors while preserving the ability to verify claims. Retain exact citations and claim-relevant excerpts so an answer can be grounded without pasting the entire retrieval corpus.

Incorrect Answers

B:

Task-specific state supports continuity without unnecessary long-term retention or cross-order disclosure; changing task-scoped memory does not establish relevant context retrieval in this application.

C:

Providing all documents defeats the stated narrow-context goal. Large irrelevant input increases token cost and can distract the model from the evidence tied to the user’s precise question.

D:

Narrow inspection reduces context consumption while preserving the option to follow dependencies when necessary; the scenario's controlling requirement is relevant context retrieval, not the related progressive contextual retrieval concern.

Question 284

A developer truncates old messages by deleting them indiscriminately. What risk does this introduce?

  1. Deleting earlier messages always improves accuracy because all old constraints are obsolete.
  2. Conversation truncation automatically preserves every referenced document version.
  3. Losing still-active constraints or unresolved task state.
  4. Only tool results, never human instructions, are vulnerable to context trimming.

Explanation

Correct Answer

C:

Context pruning must distinguish disposable observations from instructions and decisions that remain necessary for the current task.

Incorrect Answers

A:

Old history can contain still-active instructions, commitments, or unresolved requirements that must be retained.

B:

Removing messages without a summary or retrieval state can lose citation provenance and version context. Keep durable source metadata and an explicit compacted summary.

D:

Both past messages and tool observations can be lost if the application prunes them indiscriminately. A pruning rule must preserve obligations that remain in force.

Question 285

An agent retrieves twenty matching database rows but only needs the total count. What data should it receive?

  1. All matching rows must be sent so Claude can count them manually.
  2. A validated aggregate computed by the database query or tool.
  3. A complete natural-language explanation of each row is required to obtain the total.
  4. The model should infer the row count from the naming convention of the dataset.
  5. The executor should hide the query filters because only the number matters.

Explanation

Correct Answer

B:

Supplying the answer-bearing aggregate avoids spending model tokens on unnecessary row-level data. Let a database executor compute an authorized aggregate and send the count with its filtering criteria to the model.

Incorrect Answers

A:

The database or query tool can compute COUNT with consistent authorization filters and fewer prompt tokens. Return the authorized result rather than every row's private attributes.

C:

Explanations add token cost and may miscount values that a deterministic aggregate can compute exactly. Structured SQL aggregation is a more reliable computation boundary.

D:

Identifiers do not prove how many authorized rows satisfy the live query predicate. Only the query executor knows the applied live filter conditions.

E:

A count without its authorized scope can be misleading and difficult to verify in an audit. Include enough provenance for the result to be interpreted correctly.

Question 286

A support agent is grounded in a policy manual that changes weekly. What source strategy keeps answers current?

  1. Retrieve the current approved policy at query time or from a refreshed index.
  2. Fetch and include the most relevant excerpts with source identifiers.
  3. Obtain just-in-time code excerpts and expand only when evidence requires it.
  4. Persist the address in protected order-scoped state with appropriate expiry.

Explanation

Correct Answer

A:

Dynamic retrieval prevents reliance on older model knowledge or long-lived stale prompt copies. Index new approved revisions promptly and attach document version metadata so answers cite the currently effective policy.

Incorrect Answers

B:

Targeted evidence reduces input tokens and distractors while preserving the ability to verify claims; it changes relevant context retrieval; it does not prove freshness of external context in this case.

C:

Narrow inspection reduces context consumption while preserving the option to follow dependencies when necessary; the effect concerns progressive contextual retrieval; the success condition is freshness of external context.

D:

Task-specific state supports continuity without unnecessary long-term retention or cross-order disclosure; a remedy for task-scoped memory is not a substitute for freshness of external context.

Question 287

A product agent must remember the user's confirmed shipping address for this order but not across unrelated future orders. What memory scope fits?

  1. Store the address in protected order-scoped state with appropriate expiry.
  2. Obtain and include the most relevant excerpts with source identifiers.
  3. Send bounded summaries from subagents rather than full intermediate transcripts.
  4. Maintain original record references and retrieval handles alongside the concise summary.

Explanation

Correct Answer

A:

Task-specific state supports continuity without unnecessary long-term retention or cross-order disclosure. Tie the confirmed address to the order or session identifier and expire it according to the workflow's retention policy.

Incorrect Answers

B:

Targeted evidence reduces input tokens and distractors while preserving the ability to verify claims; for task-scoped memory, a control centered on relevant context retrieval is not the appropriate substitute.

C:

Context separation keeps exploratory work away from the coordinator while delivering the results it needs; this implements context isolation by delegation rather than the demanded task-scoped memory.

D:

Summaries improve context efficiency but cannot replace audit provenance when exact evidence may later be required; a remedy for source handles after compaction is not a substitute for task-scoped memory.

Question 288

A manager agent becomes distracted by thousands of subagent tool observations. What orchestration adjustment helps?

  1. As an alternative, project and validate the needed fields before adding tool_result content.
  2. Return bounded summaries from subagents rather than full intermediate transcripts.
  3. Combine an explicit current instruction hierarchy and remove superseded directions.
  4. Compact earlier turns and keep only current evidence plus durable decisions.

Explanation

Correct Answer

B:

Context separation keeps exploratory work away from the coordinator while delivering the results it needs. Have workers summarize source-linked findings, decisions, and open questions instead of forwarding all intermediate observations.

Incorrect Answers

A:

A narrow observation retains decisive evidence and reduces token consumption and prompt noise; it improves field-level tool result projection only, leaving context isolation by delegation unproven.

C:

Unresolved policy collisions create ambiguous generation behavior; the prompt must make the operative constraints identifiable; the context isolation by delegation criterion differs from conflicting instruction resolution; the gap persists.

D:

Summarization and pruning reserve attention for relevant observations; increasing context capacity alone does not repair noisy history; changing history compaction does not establish context isolation by delegation in this application.

Question 289

A tool returns a large JSON document where only three attributes affect the decision. What should the executor do?

  1. Project and validate the needed fields before adding tool_result content.
  2. Fetch the current approved policy at query time or from a refreshed index.
  3. Clear task-local history while carrying forward only deliberately selected preferences or facts.
  4. Retain original record references and retrieval handles alongside the concise summary.

Explanation

Correct Answer

A:

A narrow observation retains decisive evidence and reduces token consumption and prompt noise. Project a validated subset of fields into the tool_result and preserve a link to the original record for audit when necessary.

Incorrect Answers

B:

Dynamic retrieval prevents reliance on older model knowledge or long-lived stale prompt copies; because freshness of external context is its concern, field-level tool result projection remains outside its effect here.

C:

A new task should not inherit incidental tool traces or obsolete instructions merely because they share a session ID; the effect concerns context reset at task boundary; the success condition is field-level tool result projection.

D:

Summaries improve context efficiency but cannot replace audit provenance when exact evidence may later be required; the scenario's controlling requirement is field-level tool result projection, not the related source handles after compaction concern.

Question 290

A legal assistant must distinguish facts from prior document versions. What should its context include?

  1. Summarize earlier turns and keep only current evidence plus durable decisions.
  2. Load and include the most relevant excerpts with source identifiers.
  3. Restart task-local history while carrying forward only deliberately selected preferences or facts.
  4. Attach revision metadata and clearly label current versus superseded excerpts.
  5. Log proposed, approved, executed, and rejected statuses explicitly.

Explanation

Correct Answer

D:

Source versioning prevents the model from treating conflicting historical clauses as equally authoritative. Mark superseded clauses as historical and retrieve the currently effective text along with issue dates and jurisdiction context.

Incorrect Answers

A:

Summarization and pruning reserve attention for relevant observations; increasing context capacity alone does not repair noisy history; this targets history compaction, while the system must demonstrate version aware retrieved context.

B:

Targeted evidence reduces input tokens and distractors while preserving the ability to verify claims; it improves relevant context retrieval only, leaving version aware retrieved context unproven.

C:

A new task should not inherit incidental tool traces or obsolete instructions merely because they share a session ID; because context reset at task boundary is its concern, version aware retrieved context remains outside its effect here.

E:

Structured state prevents a narrative summary from promoting tentative discussion into authoritative instructions or completed actions; changing decision-status tracking does not establish version aware retrieved context in this application.

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