Anthropic CCAR-P Exam Dumps PDF

Question 1

A document-analysis prompt places the user's question first, followed by a 30,000-token contract. Reviewers observe that the model sometimes answers about the wrong clause. Which adjustment is most likely to improve grounding?

A. Place the long document first and the question at the end, and ask the model to quote the relevant passages before answering.
B. Increase temperature to encourage broader consideration of the document.
C. Reduce the document to its first 5,000 tokens.
D. Repeat the question five times throughout the document.

Answer: A

Question 2

A customer-facing assistant must never provide individualized financial advice, must escalate to a human when a user expresses distress, and must decline requests outside the product's scope. Where are these constraints best expressed?

A. In a post-processing filter that inspects the model's output and blocks violations.
B. In the model's training data through fine-tuning.
C. In each user message, appended by the client application.
D. In the system prompt, stated as explicit behavioral rules with defined escalation actions, reinforced by programmatic checks for the highest-risk conditions.

Answer: D

Question 3

An organization maintains eleven Claude-powered internal tools. Each has its own system prompt, and eachembeds the same 900-word block describing company tone, escalation policy, and prohibited topics. A policychange now requires editing all eleven. Which approach best addresses the maintainability problem?

A. Consolidate the eleven tools into a single application with one system prompt.
B. Shorten the shared block so that future edits are less burdensome.
C. Factor the shared block into a versioned, centrally maintained prompt module that each tool composes into its own system prompt at build or run time.
D. Instruct each tool to fetch the current policy from a database at the start of every conversation and reason about it.

Answer: C

Question 4

A model must extract structured data from semi-structured invoices that vary widely in layout. Zero-shot prompting produces correct field values but inconsistent output structure, breaking the downstream parser.Which technique most directly addresses the problem?

A. Chain-of-thought prompting, so the model reasons step by step before answering.
B. Few-shot examples demonstrating the exact output structure across several layout variations, combined with an explicitly specified output schema.
C. Raising the maximum output token limit.
D. Running the extraction twice and comparing results.

Answer: B

Question 5

A long-running agent conversation degrades in quality after roughly 40 turns: the agent begins repeating completed steps and losing track of the user's original constraints. Which context engineering approach best addresses this?

A. Introduce structured context management — periodically compact the history into a running state summary that preserves the original objective and constraints, and carry that forward instead of the full transcript.
B. Restart the conversation from scratch whenever quality degrades.
C. Instruct the agent in the system prompt to remember everything discussed.
D. Increase the temperature so the agent produces more varied responses.

Answer: A

Question 6

A team is choosing a Claude model for a high-volume, latency-sensitive intent classification step that routes messages into one of six categories. Accuracy on a held-out set is 97.8% with a smaller, faster model and 98.4% with the largest model, which costs roughly eight times more per request and adds 900 ms. Whichdecision is best supported?

A. Alternate between models to balance cost and accuracy on average.
B. Use the largest model for the first month, then re-evaluate.
C. Use the largest model, because accuracy is the primary quality metric.
D. Use the smaller model for this step, and reserve larger-model capacity for downstream steps where the accuracy delta materially changes outcomes.

Answer: D

Question 7

An application sends an identical 12,000-token system prompt and policy appendix on every request, followed by a short user message that varies. Both cost and time-to-first-token are concerns. Which change most directly improves both?

A. Switch to the smallest available model regardless of task fit.
B. Move the policy appendix into a few-shot example block at the end of the prompt.
C. Order the request so the stable system prompt and policy appendix precede all variable content, and enable prompt caching.
D. Truncate the policy appendix to its first 2,000 tokens.

Answer: C

Question 8

An architect is reviewing a proposed design in which a Claude call validates whether a submitted IBAN is well formed, checks whether a date falls within a fiscal quarter, and then assesses whether a free-text expense justification is plausible. Select TWO valid criticisms of this design.

A. Plausibility assessment of free text is unsuitable for a language model and should be handled by rules.
B. IBAN format validation is a deterministic check that should be performed in code, not by a model.
C. The design should use a larger model to improve IBAN validation accuracy.
D. Fiscal-quarter date comparison is deterministic arithmetic that should be performed in code, not by a model.
E. Combining three checks in one call is always incorrect regardless of their nature.

Answer: B,D

Question 9

A nightly job must summarize 500,000 support transcripts. There is no user waiting on the result; the only requirements are that the job completes before 6:00 a.m. and that cost is minimized. Which processing design is most appropriate?

A. Real-time synchronous calls behind an autoscaling API tier, sized for peak throughput.
B. Asynchronous batch processing with a smaller model selected by evaluation, prompt caching for the shared instruction prefix, and concurrency tuned to the completion deadline.
C. A multi-agent system that assigns each transcript to a specialist agent by product area.
D. A single request containing as many transcripts as the context window allows, repeated until the corpus is exhausted.

Answer: B

Question 10

An internal support assistant must answer questions from a 200,000-document policy library that is updated daily by dozens of authors. Answers must cite the specific policy version in force. Which approach best fits the requirement?

A. Build a retrieval pipeline over the versioned policy store, returning passages with their version metadata for the model to cite.
B. Give the model a tool that returns the full text of any policy by ID and instruct it to guess relevant IDs.
C. Fine-tune a model on the policy library and retrain nightly as documents change.
D. Place the most frequently referenced policies in the system prompt and rely on the model's general knowledge for the rest.

Answer: A