CCAO-F Practice Test 14 — 15 Questions

Question 1

A data scientist is using Claude to classify customer feedback into categories: 'bug', 'feature request', 'complaint', or 'praise'. The feedback is often short and informal. The data scientist wants to maximize classification accuracy. Which prompting strategy is most effective?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Include a few examples of feedback for each category in the prompt.. For classification tasks with informal and varied input, providing a few labeled examples per category is the most effective prompting strategy. Examples allow Claude to learn the specific patterns and boundaries between categories, improving accuracy. This few-shot approach outperforms detailed definitions alone because it demonstrates how to handle nuances like slang or brevity. It also avoids the potential confusion of step-by-step reasoning for a straightforward task.

Answer

Use a high temperature to encourage diverse interpretations of the feedback.

Answer

Provide a detailed definition of each category in the system prompt.

Answer

Ask Claude to think step-by-step about the sentiment before classifying.

Question 2

A developer is building a Claude-powered chatbot for a mental health support app. During testing, a user prompt expresses suicidal ideation. The developer wants to ensure the chatbot responds safely and appropriately. Which of the following is the best course of action according to Anthropic's safety guidelines?

Accepted Answer

Have the chatbot express empathy, encourage the user to seek professional help, and provide crisis resources without attempting to diagnose or treat.. The best action is for the chatbot to express empathy, encourage professional help, and provide crisis resources without diagnosing or treating. This aligns with Anthropic's guidelines for handling sensitive mental health topics, ensuring the user feels heard while directing them to appropriate care. It balances safety with helpfulness and avoids overstepping the chatbot's role.

Answer

Program the chatbot to offer a diagnosis of depression and suggest coping strategies.

Answer

Configure the chatbot to immediately provide a list of emergency hotlines and end the conversation.

Answer

Instruct the chatbot to change the subject to a positive topic to de-escalate the situation.

Question 3

A team is building a document Q&A feature on Claude and wants to reduce hallucinations when the answer is not present in the supplied documents. Which TWO techniques are appropriate? (Choose two.)

Accepted Answer

Instruct Claude in the system prompt to answer only from the provided documents and to say it does not know when the answer is absent.. Reducing hallucinations in document Q&A comes from constraining Claude to the supplied evidence and making that grounding verifiable. An instruction to answer only from the documents and to admit uncertainty sets the behavioral boundary, while requiring citations or quotes makes unsupported claims detectable. Sampling temperature and output-length settings do not improve factual grounding.

Answer

Remove the documents from the prompt and rely on Claude's pretrained knowledge to fill in missing details.

Answer

Increase max_tokens so Claude has more space to explain its reasoning before giving the final answer.

Answer

Raise the temperature so Claude explores a wider range of possible answers and is more likely to find the right one.

Question 4

A developer is writing an assistant that must always answer in strict JSON with a fixed set of keys. They want to guarantee the model's output is parseable without writing custom repair logic. Which approach best fits the Messages API?

Accepted Answer

Define a tool with an input schema and require the model to call it.. The Messages API supports tool use, where each tool declares an input_schema in JSON Schema. When the model is required to call that tool, the returned arguments conform to the declared schema, yielding reliably structured output. Sampling settings, stop sequences, and prompt wording all shape behavior probabilistically but cannot guarantee a parseable, fixed-key JSON object.

Answer

Set 'temperature' to 0 and rely on the model to output valid JSON.

Answer

Append the word 'JSON' to the system prompt and trust the model's compliance.

Answer

Use the 'stop_sequences' parameter to halt generation at the closing brace.

Question 5

A developer is designing a Claude-powered assistant that must maintain a coherent conversation across many turns while controlling cost and context limits. Which TWO practices are appropriate according to Claude model fundamentals? (Choose two.)

Accepted Answer

Summarize or truncate older turns to keep the context within limits while preserving recent, relevant exchange.. Because the Messages API is stateless, long conversations require deliberate context management. Summarizing or truncating older turns keeps requests within the context window and controls cost while preserving recent continuity. Durable persona, tone, and formatting rules belong in the system prompt, where they persist across turns without repetition. Relying on automatic memory or misusing max_tokens misunderstands how Claude processes each independent request.

Answer

Increase max_tokens on every request to allow Claude to store more conversation history internally.

Answer

Rely on Claude to remember previous turns automatically between separate API calls.

Answer

Send the full conversation history with each request so Claude has complete context.

Question 6

A support team wants Claude to classify incoming customer emails into one of five categories: Billing, Technical, Account, Feature Request, or Other. The team needs consistent, machine-readable output that a downstream script can parse reliably. Which prompt design best meets this requirement?

Accepted Answer

Instruct Claude to output only the category label, chosen from the predefined list, with no additional text.. When output feeds an automated script, the prompt should constrain the response to a single value from a predefined set and forbid extra text. This minimizes parsing complexity and ambiguity. Few-shot examples can improve consistency, but adding confidence scores or allowing multiple labels reintroduces variability. Narrative explanations with formatting are human-friendly but not reliably machine-readable, so they fail the stated integration requirement.

Answer

Ask Claude to choose the best category and, if multiple categories seem to apply, list them all separated by commas.

Answer

Ask Claude to explain its reasoning in a paragraph and end with the category name in bold.

Answer

Provide three examples of emails and their categories, then ask Claude to respond with the category and a confidence score between 0 and 100.

Question 7

A customer support team is building a Claude-powered assistant that must answer questions using a 300-page product manual. They want to avoid sending the entire manual with every request because of latency and cost. Which approach best leverages Claude's capabilities while keeping responses grounded in the manual?

Accepted Answer

Use retrieval-augmented generation: embed the manual, retrieve the most relevant passages, and include them in the prompt.. Retrieval-augmented generation is the standard architecture for grounding Claude in a large, private, frequently updated document such as a product manual. Embedding the manual and retrieving only the most relevant passages keeps prompts small, reduces latency and cost, and gives Claude authoritative source text to answer from. Fine-tuning, max_tokens, and temperature do not supply the missing manual content and therefore cannot ensure grounded support answers.

Answer

Lower the temperature to 0 so Claude reproduces the manual verbatim from memory.

Answer

Increase the max_tokens parameter so Claude can generate longer, more detailed answers directly from its training data.

Answer

Fine-tune Claude on the manual so the knowledge is embedded in the model weights.

Question 8

A developer is writing code that calls the Anthropic Messages API and needs to authenticate each request. The developer has retrieved the API key from a secure secret manager at runtime. Where should the API key be placed in the HTTP request?

Accepted Answer

In the 'x-api-key' HTTP header on each request.. Direct HTTP calls to the Anthropic Messages API authenticate by including the API key in the x-api-key request header. This keeps the credential out of URLs and payloads, which is important for security and log hygiene. Body fields, query parameters, and Basic authentication schemes are not recognized by the API for this purpose, so requests using them will fail authentication.

Answer

In the URL as a query parameter such as ?api_key=sk-ant-...

Answer

In the request body as a top-level 'api_key' field alongside 'model' and 'messages'.

Answer

In the 'Authorization' header using the scheme 'Basic' with the key base64-encoded.

Question 9

A product manager is drafting a system prompt for a customer-facing Claude assistant. They want the assistant to maintain a professional tone, avoid discussing competitors, and always end responses with a link to the help center. They also want to allow users to override the tone for casual conversations. Which statement about system prompts best guides this design?

Accepted Answer

The system prompt should set the global role and rules, and the assistant can be instructed to allow user requests to adjust tone within defined bounds.. System prompts are the appropriate place for persistent, application-controlled behavior such as role, tone, and global restrictions, and they can also specify how user requests may modify certain aspects within defined limits. This provides consistent guardrails while allowing the desired flexibility. Relying on user messages, leaving the system prompt empty, or forbidding all overrides either weakens the guardrails or removes the intended flexibility.

Answer

The system prompt should be left empty, and all rules including tone and competitor avoidance should be embedded in the first user message of every session.

Answer

The system prompt should instruct Claude to ignore all user attempts to change its behavior, ensuring the professional tone and competitor rules are never overridden.

Answer

The system prompt should contain only the tone and competitor rules, while the help center link instruction belongs in each user turn to ensure it is followed.

Question 10

A developer is building a Claude-based assistant that must answer questions about a 90,000-token product manual. The manual is provided in the prompt on every request. The developer wants to improve answer accuracy and reduce the chance of the model overlooking relevant sections. Which TWO techniques are most appropriate? (Choose two.)

Accepted Answer

Instruct Claude to answer only from the provided manual and to say 'I don't know' if the answer is not present.. Long-context accuracy improves when the source material is placed before the question and instructions, and when the model is explicitly told to answer only from that source with a not-found fallback. These two techniques together reduce overlooking and hallucination. Raising temperature, summarizing first, or randomly sampling fragments either adds noise, loses detail, or risks omitting the relevant content entirely.

Answer

Ask Claude to summarize the entire manual first, then answer the question based only on that summary.

Answer

Increase the temperature to 1.0 so Claude considers a wider range of interpretations from the manual.

Answer

Split the manual into random fragments and include only a few fragments per request to keep the prompt short.

Question 11

A developer is preparing to deploy Claude in an application that generates creative fiction. They want to ensure the application is used responsibly and complies with Anthropic's Usage Policies. Which TWO practices should they implement? (Choose two.)

Accepted Answer

Include a clear disclosure to end users that content is AI-generated.. Disclosing AI-generated content and providing a reporting mechanism are two concrete practices that support transparency and accountability. They align with responsible use expectations for creative applications without unnecessarily restricting legitimate expression. The other options are either overly broad, ineffective, or not required by policy.

Answer

Require users to sign a legal waiver acknowledging that AI-generated content may be offensive.

Answer

Fine-tune Claude to never generate content that could be considered controversial.

Answer

Implement a content filter to block any violent or sexual themes in the generated fiction.

Question 12

A developer wants Claude to call an internal inventory function when a user asks about stock levels, but the function must only be invoked when the model decides it is needed. The application will execute the function and return the result. Which sequence correctly implements this with the Messages API?

Accepted Answer

Send the tools definition; if the response contains a tool_use block, run the function and send a new request including a tool_result block referencing the tool_use id.. Tool use is a round trip: the application declares tools, the model optionally returns a tool_use block, the application executes the named function, and it sends back a tool_result block that references the tool_use id so the model can incorporate the real output. Pre-running functions, prose-based parsing, or empty acknowledgements all break this contract.

Answer

Pre-execute the inventory function on every request and inject its output into the system prompt.

Answer

Define the function in the system prompt as a description and parse the model's prose reply for a function name.

Answer

Send the tools definition and, upon receiving a tool_use block, immediately send a tool_result with an empty payload to acknowledge it.

Question 13

An engineer is choosing between Claude 3.5 Sonnet and Claude 3 Opus for a pipeline that extracts structured fields from scanned invoices. The pipeline processes thousands of documents per hour and must balance accuracy against cost. Which statement best reflects the appropriate model-selection reasoning?

Accepted Answer

Claude 3.5 Sonnet is often the better balance because it provides strong accuracy on structured extraction at lower cost and higher throughput than Opus.. Model selection should match capability to task and volume. Claude 3.5 Sonnet offers strong structured-extraction accuracy at a lower price and higher throughput than Claude 3 Opus, making it the sensible default for a high-volume invoice pipeline. Reserving Opus for the hardest cases balances quality and cost. Treating all Claude 3 models as identical, or assuming runtime knowledge transfer between models, misstates how the API and model tiers actually work.

Answer

Claude 3 Opus should be used only for the first document and Sonnet for the rest, because Opus can teach Sonnet the schema at runtime.

Answer

Claude 3 Opus should always be chosen because it is the most capable model and therefore the most cost-effective at scale.

Answer

Model choice is irrelevant because all Claude 3 models have identical pricing and performance characteristics.

Question 14

A product team drafts a system prompt for a customer-facing assistant. The draft is 4,000 words and mixes persona description, tone guidance, formatting rules, escalation policy, and several anecdotes about past incidents. Reviewers find that Claude follows the formatting rules but frequently ignores the escalation policy. Which change best improves adherence to the escalation policy?

Accepted Answer

Restructure the system prompt into clearly labeled sections and place the escalation policy in its own section with explicit trigger conditions.. The escalation policy is being lost among unrelated narrative content. Giving it a dedicated, labeled section with explicit trigger conditions turns it into an unambiguous, retrievable instruction. Structural organization reduces interference between competing directives, which is why the fix is about clarity and grouping rather than length alone or ordering alone.

Answer

Shorten the system prompt by deleting the formatting rules so the escalation policy receives more attention.

Answer

Convert the entire system prompt into a numbered list of every rule in the order it was originally written.

Answer

Move the escalation policy to the last line of the system prompt without changing its wording.

Question 15

A researcher is using Claude to analyze sensitive personal data from a study. They want to ensure the data is handled responsibly and in line with Anthropic's guidelines. Which action is most appropriate before sending the data to Claude?

Accepted Answer

Anonymize or de-identify the personal data before including it in the prompt.. Anonymizing or de-identifying personal data before sending it to Claude is the most direct way to protect privacy and comply with responsible data handling guidelines. Encryption without decryption, consent alone, or fine-tuning on sensitive data do not adequately address the risk. De-identification allows analysis while minimizing exposure.

Answer

Encrypt the data and send it as a base64-encoded string in the prompt.

Answer

Obtain verbal consent from the study participants to use their data with an AI system.

Answer

Use a custom fine-tuned model that has been trained on similar sensitive data.