CCDV-F Practice Test 5 — 25 Questions

Question 1

A developer is writing a support-triage prompt for Claude. The prompt contains a 12-page product manual followed by the customer's question. Testing shows Claude sometimes answers using general knowledge about competing products instead of the manual. The developer wants Claude to ground every answer in the manual and to say 'not covered' when the manual lacks the answer. Which change best achieves this?

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

Accepted Answer

Instruct Claude to quote the relevant manual passage before answering, and to reply 'not covered' if no passage supports the answer.. Grounding improves when the model must produce evidence before its conclusion, because the quoted passage anchors the answer in the manual. Giving an explicit 'not covered' response removes the pressure to invent an answer, which is what allows general knowledge to leak in. Together these two elements address both the sourcing and the fallback behavior.

Answer

Shorten the manual to its first two pages so the context is easier for Claude to process.

Answer

Move the customer question to the top of the prompt so Claude reads the question before the manual.

Answer

Add the sentence 'Be accurate and do not hallucinate' to the end of the prompt.

Question 2

A developer is building a Claude agent with the Anthropic SDK that must look up live order status. The agent has one tool named get_order_status, and its input_schema declares a required string property order_id. During testing, Claude returns a tool_use block with name "get_order_status" and input {"order_id": "A-4471"}. What must the developer's application do next to keep the agent loop running correctly?

Accepted Answer

Execute the lookup, then send a user-role message containing a tool_result block with the matching tool_use_id back to Claude.. The agent loop alternates model reasoning with client-side tool execution. When Claude emits a tool_use block, the developer's code runs the named tool and returns the output as a tool_result content block in a user message, keyed by the same tool_use_id. That correlation lets Claude pair the result with its request and produce the final answer using real order data.

Answer

Call the tool, then send the raw JSON output as a plain assistant text message so Claude can read the order details directly.

Answer

Append Claude's tool_use block to the conversation and immediately call the Messages API again with the same parameters to obtain the final answer.

Answer

Convert the tool_use block into a system message so the model treats the tool request as persistent instructions for the rest of the session.

Question 3

A developer is working on a large refactoring task with Claude Code. They want to ensure the agent has sufficient context about the project's architecture and coding standards. Which TWO of the following are recommended ways to provide this context? (Choose two.)

Accepted Answer

During the session, explicitly reference key files like `src/core/engine.ts` and ask Claude to read them for context.. The recommended ways to provide project context are using a `CLAUDE.md` file at the root and explicitly referencing key files during the session. `CLAUDE.md` offers persistent, project-wide guidance, while on-demand file references allow the agent to focus on specific areas. Other options like `/context` command, `--add-context` flag, or relying solely on comments are not supported or effective.

Answer

Run `claude --add-context architecture.md` to load the architecture document into the session.

Answer

Include detailed comments in each source file describing the overall system design.

Answer

Use the `/context` command to manually upload a PDF of the architecture diagram.

Question 4

A developer wants Claude Code to remember a project-specific convention: all new database migration files must be placed in the `db/migrations` directory and named with a `YYYYMMDDHHMMSS_` prefix. The developer wants this rule applied automatically in every future Claude Code session for this repository, without having to repeat the instruction each time. What is the most appropriate way to achieve this?

Accepted Answer

Add the convention to a CLAUDE.md file at the repository root.. Persistent, repository-scoped instructions belong in CLAUDE.md, which Claude Code reads automatically at session start. That file can be committed so every developer and every future session inherits the naming and placement rule without manual repetition. Ad-hoc pasting or non-existent config keys do not satisfy the requirement of automatic, durable application across sessions.

Answer

Run `claude config set migration.dir db/migrations` before each session.

Answer

Create a `.claudeignore` file listing `db/migrations` to protect the directory.

Answer

Paste the convention into the first message of every new session.

Question 5

A developer is using the Anthropic Messages API and wants to implement a retry mechanism for handling rate limit errors. They receive an HTTP 429 response with a 'retry-after' header. Which approach is the most appropriate for handling this error?

Accepted Answer

Wait for the number of seconds specified in the 'retry-after' header before retrying the request.. When the API returns a 429 status with a 'retry-after' header, the best practice is to pause and wait for the specified number of seconds before retrying. This respects the server's rate limiting policy and avoids making unnecessary requests that could lead to further throttling. Immediate retries or switching models do not address the root cause. Failing without retry is also not ideal for transient rate limits.

Answer

Immediately retry the request in a tight loop until it succeeds.

Answer

Switch to a different model that has higher rate limits and retry immediately.

Answer

Log the error and fail the request without retrying, as rate limits are permanent.

Question 6

A developer is using Claude to review a 180-page contract. The model must cite the exact page number for each risk it flags. Placing the entire contract in the prompt causes the model to cite vaguely or omit page numbers. Which change best improves citation accuracy without exceeding the context window?

Accepted Answer

Insert explicit page-marker tags such as '[[PAGE 42]]' before each page's text and instruct the model to cite the nearest marker.. Embedding page markers makes citation a local pattern-matching task instead of a global spatial inference. The model can still read the whole contract in one pass, preserving cross-reference reasoning, while the nearest marker gives an unambiguous page label. Temperature changes and summary-then-cite pipelines do not supply the missing positional signal.

Answer

Ask the model to summarize each page first, then cite the summary.

Answer

Lower the temperature to 0 so the model copies page numbers verbatim.

Answer

Split the contract into 180 separate API calls, one per page.

Question 7

A developer is building a document Q&A service on the Anthropic Messages API. Their integration currently sends the entire 240,000-token knowledge base on every turn of a long conversation, and they are hitting the model's context window limit. They want to keep the full conversation history and the knowledge base available without exceeding the window. Which approach best addresses the problem?

Accepted Answer

Summarize or truncate older conversation turns and retrieve only the most relevant knowledge-base chunks to include in each request.. The context window is a fixed ceiling on combined input and output tokens per request. To keep a long conversation and a large corpus usable, the developer must shrink what is actually sent: condense prior turns and inject only the retrieved passages relevant to the current question. Caching and streaming change cost or delivery mechanics, not the window itself, and max_tokens governs output only.

Answer

Enable prompt caching on the static knowledge base so repeated prefixes are served from cache and no longer count toward the context window.

Answer

Increase the max_tokens parameter so Claude can internally compress the knowledge base before answering.

Answer

Switch the request to stream: true so the server transmits the knowledge base in smaller chunks that bypass the context limit.

Question 8

A developer is using the Anthropic Messages API and wants to ensure that Claude's response is deterministic and reproducible for a given prompt. They set the temperature parameter to 0. However, they observe that repeated calls with the same input sometimes yield slightly different outputs. Which factor is the most likely cause of this non-determinism?

Accepted Answer

The model's internal computations involve non-deterministic operations such as floating-point arithmetic or parallel processing.. Even with temperature set to 0, which makes the model choose the most probable next token, the underlying probability distribution can vary slightly between calls due to non-deterministic operations in the model's execution on distributed hardware. This includes floating-point arithmetic order and parallel processing. The API does not provide a seed parameter to enforce determinism. Therefore, the most likely cause is the model's internal non-determinism, not misconfiguration of temperature or top_p.

Answer

The API automatically injects a random seed into each request unless one is explicitly provided.

Answer

The top_p parameter is set to a value less than 1, introducing randomness even with temperature 0.

Answer

The temperature parameter is ignored when set to 0, and the default temperature is used instead.

Question 9

A team is estimating costs for a new feature that will send 1 million requests per month. Each request has a 2,000-token input and generates a 500-token output. The team wants to reduce the output token cost, which dominates the bill. Which strategy is MOST effective for reducing output token costs?

Accepted Answer

Instruct the model to be concise and set a lower max_tokens limit to cap response length.. Because output tokens dominate the monthly bill, the most effective lever is limiting how many tokens the model generates. Lowering max_tokens and prompting for conciseness directly caps output length and thus cost. Optimizing input tokens or context capacity addresses smaller or irrelevant components of the bill.

Answer

Enable Prompt Caching on the input to reduce the input token cost.

Answer

Increase the max_tokens parameter to allow longer responses.

Answer

Switch to a model with a larger context window to accommodate longer inputs.

Question 10

A developer is building an MCP client that connects to a remote MCP server over HTTP with Server-Sent Events. The server advertises tools that require user-specific permissions. Which TWO practices should the developer implement to keep the integration secure and functional? (Choose two.)

Accepted Answer

Validate the server's identity and TLS certificate, and reject connections that fail certificate verification.. Remote MCP servers with permission-scoped tools require two things: proof of who the user is, and protection of the channel carrying that proof. Completing the server's OAuth flow and sending the access token satisfies authentication, while validating TLS certificates prevents interception or impersonation. Shared credentials, plaintext transport, and indefinite caching all undermine per-user authorization and would make the integration both insecure and unreliable.

Answer

Hard-code a shared service account token in the client binary so all users inherit the same permissions.

Answer

Disable TLS termination to reduce latency, since MCP messages are already structured JSON.

Answer

Cache all tool results on the client indefinitely so users avoid repeated permission checks.

Question 11

A developer is building a real-time translation service that must respond within 1 second for short phrases. The service will handle thousands of requests per hour. Quality is important, but latency and cost are critical. Which Claude model is the most appropriate?

Accepted Answer

Claude 3 Haiku. Claude 3 Haiku is optimized for low latency and low cost, making it the best fit for real-time translation of short phrases. It can meet the 1-second response requirement and handle high throughput without excessive expense. More powerful models like Sonnet or Opus would add latency and cost without necessary quality gains for this straightforward task.

Answer

Claude 3 Opus

Answer

Claude 3.5 Sonnet

Answer

Claude 3 Sonnet

Question 12

A developer is building a Claude-powered agent that uses the Anthropic API with a tool-use loop. The agent can invoke a `fetch_url` tool that retrieves the contents of any URL supplied by the model. During a red-team exercise, an attacker embeds hidden instructions in a page the agent fetches, causing the agent to call `fetch_url` again with an attacker-controlled URL containing sensitive query parameters. Which TWO controls best reduce this tool-use loop risk? (Choose two.)

Accepted Answer

Restrict the `fetch_url` tool to an allowlist of approved domains and schemes, and reject any URL not on that list before the tool executes.. The attack works because fetched content is untrusted and can carry instructions that the agent treats as goals. Effective defense combines limiting where the tool can go with treating tool output as data rather than commands and requiring human approval for outbound calls. These controls stop the second-stage request before it can carry sensitive parameters to an attacker-controlled endpoint.

Answer

Log every tool invocation with its arguments and retain the logs for forensic analysis after an incident.

Answer

Increase the model's temperature setting so it is less likely to follow deterministic hidden instructions embedded in fetched content.

Answer

Raise the `max_tokens` value on each API request so the agent has enough context to distinguish legitimate instructions from injected ones.

Question 13

A developer is using the Anthropic SDK to build an agent that can answer questions by calling external APIs. The agent's tool returns a large JSON payload (over 100,000 tokens) as a `tool_result`. The developer notices that subsequent requests to the model fail with a context window error. What is the most effective way to resolve this issue?

Accepted Answer

Truncate or summarize the tool result before including it in the conversation history.. The most effective solution is to truncate or summarize the tool result before adding it to the conversation history. The context window includes all input tokens, so a massive tool result will cause errors. By reducing the size, the developer ensures the conversation stays within limits. Other options do not address the root cause of the oversized input.

Answer

Increase the `max_tokens` parameter in the request to accommodate the large tool result.

Answer

Set the `tool_choice` parameter to `{"type": "none"}` to prevent the model from calling the tool again.

Answer

Switch to a model with a larger context window, such as Claude 3.5 Sonnet, without modifying the tool result.

Question 14

A developer is comparing the cost of using Claude 3 Opus versus Claude 3.5 Sonnet for a task that requires complex reasoning. The task involves processing 1,000 requests, each with 500 input tokens and 200 output tokens. Which statement accurately reflects the cost consideration?

Accepted Answer

Claude 3.5 Sonnet is less expensive per token than Claude 3 Opus, but it may require more tokens to match Opus's reasoning quality.. When choosing between Claude 3 Opus and Claude 3.5 Sonnet for complex reasoning, developers must weigh per-token cost against the number of tokens needed. Sonnet is cheaper per token, but Opus may deliver better results with fewer tokens. The correct statement acknowledges that Sonnet is less expensive per token but might require more tokens to match Opus's quality, making total cost dependent on the specific task and prompt engineering.

Answer

The cost is identical because both models are billed at the same rate for input and output tokens.

Answer

Claude 3 Opus is cheaper for output tokens but more expensive for input tokens compared to Claude 3.5 Sonnet.

Answer

Claude 3 Opus is always more cost-effective for complex reasoning because it requires fewer tokens to achieve the same result.

Question 15

A developer needs Claude to transform a list of product descriptions into a fixed XML schema that a downstream parser expects. The model sometimes adds a friendly introductory sentence before the XML. Which change most directly eliminates the extra prose?

Accepted Answer

Add a clear instruction that the response must begin immediately with the opening XML tag and contain nothing else.. Explicit output-format constraints are the direct fix for unwanted preamble. Telling the model exactly where the response must begin, and that nothing may precede it, removes the ambiguity that lets conversational framing leak in. Temperature, input length, and added reasoning all fail to target the specific behavior.

Answer

Shorten the input descriptions so the model has less to respond to.

Answer

Increase temperature so the model varies its phrasing and may omit the introduction.

Answer

Ask the model to explain each transformation step before producing the XML.

Question 16

A developer is building a Claude-powered assistant that uses a tool called 'search_knowledge_base'. The tool definition includes a required parameter 'query' and an optional 'max_results' parameter. During testing, Claude sometimes calls the tool without providing 'max_results'. What should the developer do to ensure the tool call is handled correctly?

Accepted Answer

In the tool implementation, check if 'max_results' is present and apply a sensible default value if it is missing.. Because 'max_results' is optional, Claude may omit it. The tool implementation should check for its presence and apply a default value, ensuring the tool works correctly regardless. This approach maintains the intended flexibility of the tool definition and prevents runtime errors. Other options either incorrectly change the tool contract or rely on unreliable prompting.

Answer

Add a description to the tool definition instructing Claude to always include 'max_results' with a value of 10.

Answer

Change the tool to accept a single string parameter that combines the query and max_results, such as 'query|max_results'.

Answer

Make 'max_results' a required parameter in the tool definition so Claude must always provide it.

Question 17

A developer is building an agent using the Anthropic Python SDK. The agent's system prompt instructs it to answer questions about a company's internal policies by searching a vector database. The developer wants the agent to autonomously decide when to search and when to answer directly. Which combination of API features should the developer implement to achieve this?

Accepted Answer

Define a tool that queries the vector database, include it in the `tools` parameter of the Messages API request, and handle `tool_use` blocks in the response by executing the search and returning a `tool_result` block.. The correct approach is to define the vector search as a tool, include it in the request, and handle the tool_use/tool_result cycle. This enables the model to decide when to search based on the conversation, which is the core of agentic behavior with the Anthropic SDK. Other options either force tool use, embed excessive context, or bypass the model's decision-making entirely.

Answer

Use the `system` parameter to embed the entire vector database content as context, allowing the model to answer directly without tool use.

Answer

Implement a pre-processing step that always queries the vector database and appends the top result to the user's message before sending it to the model.

Answer

Set the `tool_choice` parameter to `{"type": "any"}` so the model always calls the vector search tool, and then parse the result to generate the final answer.

Question 18

A developer is building a customer-support bot with the Claude Messages API. The system prompt instructs Claude to answer concisely, but the bot also needs to return a stable JSON object with the fields 'intent', 'sentiment', and 'reply' on every turn. Where should the instruction to produce this JSON object be placed to most reliably control the response format?

Accepted Answer

In the system prompt, placed after the conciseness instruction.. The system prompt is the correct place because it is sent with every Messages API request and reliably conditions all assistant turns. Interface requirements such as JSON keys should live there so they are not diluted by user content. Putting format rules in the user turn or in a lone assistant example makes them fragile and easy to override.

Answer

Only in the first assistant message as an example.

Answer

Appended to each user message as a trailing reminder.

Answer

In a separate follow-up call after Claude replies in prose.

Question 19

An agent built with the Claude Agent SDK runs a long research task and repeatedly hits the model's context window limit. The developer wants the agent to keep working without losing critical earlier findings. Which approach best addresses this?

Accepted Answer

Enable automatic context compaction so older turns are summarized into a condensed form while key findings are retained in the running context.. Long-running agents need a strategy for the finite context window. Compaction condenses older turns into summaries or retained highlights, freeing tokens while keeping the findings the agent depends on. Sampling parameters and response-length caps do not reclaim history, and restarting loses state, so compaction is the mechanism that lets the run continue coherently.

Answer

Raise the temperature so the model generates shorter responses and consumes fewer tokens per turn.

Answer

Restart the agent from scratch each time the limit is reached, relying on the model to re-derive earlier findings from the task prompt.

Answer

Increase the model's max_tokens parameter so each response can be longer and the context window is used more efficiently.

Question 20

A developer is writing a tool called `send_email` for a Claude-based assistant. The tool's description currently reads: 'Sends an email.' During testing, Claude calls the tool for tasks like drafting an email or checking inbox contents. Which revision to the tool definition most directly improves Claude's decision about when to invoke `send_email`?

Accepted Answer

Rewrite the description to state precisely what the tool does, when to use it, and when not to use it, such as only when the user explicitly asks to send a message.. The model chooses tools primarily from their descriptions, so a vague one like 'Sends an email' invites calls for drafting or inbox checks. A description that states the exact action, the conditions that should trigger it, and explicit non-triggers such as drafting or retrieving messages gives Claude the criteria it needs. Schema changes and renaming affect argument validity or naming, not the decision of when the tool applies.

Answer

Set `input_schema` to `type: "object"` with `additionalProperties: false` to lock down the parameter set.

Answer

Rename the tool from `send_email` to `email` and rely on the shorter name to signal its broader role.

Answer

Add a `required` array listing `to`, `subject`, and `body` in the input schema.