CCDV-F Practice Exam 3 — 20 Questions

Question 1

A developer is building a Claude-based agent that uses tool calling to interact with a database. During testing, the agent executes a tool that deletes records when the user's prompt contains the phrase 'clean up old data'. The developer wants to prevent unintended destructive actions while still allowing the agent to propose deletions. Which design change is most effective?

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

Accepted Answer

Implement a human-in-the-loop approval step where any tool call that deletes data requires explicit user confirmation before execution.. The most effective change is to add a human-in-the-loop approval step for destructive tool calls. This enforces a hard control that prevents unintended deletions regardless of model behavior, while still allowing the agent to propose actions and a human to confirm them.

Answer

Use a more powerful model with better reasoning to avoid misinterpreting the prompt.

Answer

Limit the database user's permissions to read-only, so the delete tool will fail if called.

Answer

Add a system prompt instruction telling Claude to never delete data without explicit user confirmation.

Question 2

You are building a support-triage assistant on the Anthropic API. Each request must return a JSON object with keys 'category' and 'priority'. During testing, Claude wraps its output in markdown fences and adds a friendly sentence before the JSON. You want the raw, parseable object every time without changing the model or adding a second call. What is the most reliable prompt-level change?

Accepted Answer

Pre-fill the assistant turn with an opening brace so the response must continue as the JSON object.. Constraining the assistant turn with a leading brace forces continuation as the JSON object, since the model must complete the already-started structure. This removes markdown fences and conversational preamble in a single call, with no model change. Instruction-only or explanation-first approaches leave room for stray text, so prefilling is the most reliable prompt-level fix for guaranteed parseable output.

Answer

Ask Claude to explain its reasoning first, then append the JSON at the end of the response.

Answer

Add the sentence 'Please output only JSON' to the system prompt and rely on that instruction alone.

Answer

Increase the temperature setting so the model has more freedom to choose a clean format.

Question 3

A developer is defining a tool for Claude that fetches the current stock price for a ticker symbol. The tool will be called `get_stock_price`. Which `input_schema` definition best enables Claude to call the tool correctly?

Accepted Answer

A JSON Schema object with `type: "object"`, a `properties` entry for `ticker` of `type: "string"`, and a `required` array containing `"ticker"`.. Tool definitions should describe inputs as a JSON Schema object with named, typed properties and a `required` array. For a stock lookup, a required string `ticker` tells Claude exactly what to supply. Permissive schemas, non-object root types, or missing required/description fields all degrade the model's ability to produce correct tool call arguments, so the fully specified object schema is the right choice.

Answer

A JSON Schema object with `type: "object"` and a free-form `additionalProperties: true` allowing any fields.

Answer

A JSON Schema object with `type: "object"` and a `properties` entry for `ticker`, but no `required` array and no `description`.

Answer

A JSON Schema object with `type: "string"` and a `description` explaining that the value should be a ticker symbol.

Question 4

A developer is implementing a multi-agent system using the Anthropic SDK where a 'Coordinator' agent delegates tasks to specialized 'Researcher' and 'Writer' agents via tool calls. The Coordinator must maintain a coherent conversation state and avoid redundant work. Which TWO of the following practices are essential for the Coordinator to effectively manage this delegation? (Choose two.)

Accepted Answer

Include the full conversation history, including all tool_use and tool_result blocks, in each subsequent request to the Coordinator so it can reason about prior delegations.. Effective multi-agent coordination with the Anthropic SDK requires maintaining the full conversation history to provide context, and defining clear, well-described tools for each specialized agent. These practices enable the Coordinator to make informed delegation decisions and avoid redundancy. Other options are either not essential or potentially harmful.

Answer

Set the `temperature` to 0 for the Coordinator to ensure deterministic routing decisions.

Answer

Implement a separate memory store (e.g., a database) that the Coordinator queries before each delegation to check if the task was already completed.

Answer

Use the `max_tokens` parameter to limit the Coordinator's response length, ensuring it does not generate overly long delegations.

Question 5

A developer is designing an agent with the Anthropic SDK that must follow strict brand guidelines when drafting customer emails. The agent should never deviate from these rules regardless of how the user phrases a request. Where should the brand guidelines be placed to have the strongest effect?

Accepted Answer

In the system prompt, so they act as persistent instructions across the entire conversation.. The system prompt is the correct place for durable rules that must apply to every response, such as brand guidelines. It carries more weight than user turns and persists across the conversation, making it the strongest lever for preventing deviations regardless of how a request is phrased.

Answer

In the assistant's previous reply, so the model imitates its own earlier tone.

Answer

In the first user message, so the model sees them before any other request.

Answer

In a tool description, so the model reads them whenever it considers using a tool.

Question 6

A developer wants Claude to return a fixed set of user-profile fields extracted from free-text bios. They want the response to be valid JSON they can parse without regex. The model occasionally adds a friendly sentence before the JSON. Which Messages API feature most reliably constrains the output?

Accepted Answer

Define a tool whose input_schema describes the profile fields, then set tool_choice to force Claude to call it.. Structured extraction is best enforced by defining a tool whose input_schema specifies the exact fields and types, then using tool_choice to require that the model call it. Claude then returns a tool_use block whose input is JSON matching the schema, eliminating stray prose. Prompt wording, temperature, stop sequences, and token limits influence style or length but cannot enforce a schema.

Answer

Increase max_tokens so Claude has more room to format the JSON correctly.

Answer

Set stop_sequences to "}" so generation halts exactly at the end of the JSON object.

Answer

Add the word "JSON" to the prompt and set temperature to 0.

Question 7

A team is building a retrieval-augmented assistant that sends a large document set inside the system prompt on every turn. To reduce cost, they enable prompt caching. They notice cache_read_input_tokens is high on most turns but intermittently drops to zero even though the system prompt text has not changed. Which explanation best fits this behavior?

Accepted Answer

The cache has a short time-to-live, so if enough time passes between requests the cached prefix expires and the next request pays full input cost before re-populating the cache.. Prompt caching entries expire after a time-to-live if they are not referenced. Bursty or idle traffic patterns cause the prefix to be evicted between requests, so the next call misses the cache and reports zero cache reads before re-creating the entry. Keeping steady traffic, or reducing gaps between calls, stabilizes cache_read_input_tokens and preserves the intended cost savings.

Answer

The API randomly invalidates cache entries per account to distribute load, so intermittent cache misses are expected and cannot be controlled.

Answer

Prompt caching only applies to user messages, so any content placed in the system prompt bypasses the cache and is billed at full input rate.

Answer

Cache reads only register when max_tokens is large enough to cover the entire cached prefix, so smaller responses skip the cache.

Question 8

A developer is building a support agent with the Claude Agent SDK. The agent must pause execution, ask a human operator to approve a refund above $500, and resume the exact same session with the operator's decision injected as a new message. Which SDK capability should the developer configure to accomplish this reliably?

Accepted Answer

Permission callbacks combined with session persistence, so the agent suspends on the sensitive action and later resumes from the saved session state.. Approval gates require an enforcement point plus durable state. Permission callbacks intercept the sensitive tool call before it executes, and session persistence stores the conversation so the run can be resumed later with the operator's decision appended. Together they suspend and continue the same session deterministically, which prompt instructions, blocking handlers, or retry loops cannot provide.

Answer

Wrapping the refund tool in a retry loop that re-invokes the model until the operator approves in a separate chat window.

Answer

A custom tool whose handler blocks on stdin and returns the operator's typed answer as a tool_result.

Answer

Increasing max_tokens and instructing the model in the system prompt to ask for confirmation before calling the refund tool.

Question 9

A developer is building a customer-support agent using the Anthropic API with the Claude 3.5 Sonnet model. The agent exposes a `lookup_order` tool that takes an `order_id` string. During testing, a user asks 'Where is my order 88213?' and Claude returns a `tool_use` block with `name: "lookup_order"` and `input: {"order_id": "88213"}`. What must the developer send back to Claude in the next API request so it can produce a natural-language answer?

Accepted Answer

A `user` message containing a `tool_result` content block whose `tool_use_id` matches the `tool_use` block and whose `content` holds the lookup result.. After Claude emits a `tool_use` block, the developer must return a user message containing a `tool_result` block that carries the same `tool_use_id`. This pairing is how the model links its request to the returned data. Once the matching result is provided, Claude can generate the final answer about the order. Any other message shape leaves the tool call unsatisfied and prevents a grounded response.

Answer

A new `system` message that contains the order details and instructs Claude to answer the user.

Answer

An `assistant` message that echoes the `tool_use` block and appends the lookup result inside the same block.

Answer

A `user` message containing the raw JSON result of the order lookup, with no `tool_use_id` reference.

Question 10

A developer's agent calls a weather API tool but Claude keeps emitting the tool call with a misspelled parameter name that the API rejects. The tool definition and the API schema disagree. What is the most likely cause?

Accepted Answer

The tool's input_schema does not accurately describe the parameter names and types the handler expects.. Tool inputs are generated from the declared input_schema, so any divergence between that schema and the handler's real expectations surfaces as malformed arguments. The model is not guessing; it is following the contract it was given. Correcting the schema to match the API and handler resolves the misspelled or mistyped parameter problem at its source.

Answer

The model's temperature is set too low, causing it to repeat an earlier mistake.

Answer

The system prompt does not explicitly forbid the model from inventing parameter names.

Answer

The tool_result returned to the model is too long, so the model forgets the correct parameter names.

Question 11

A developer builds a retrieval-augmented generation (RAG) system where Claude answers questions using documents stored in an internal knowledge base. The security team is concerned that a malicious document could contain instructions that hijack the model. Which design choice most effectively reduces this indirect prompt injection risk?

Accepted Answer

Treat retrieved document content as data only, place it in a clearly delimited user turn, and instruct the model to never follow instructions found inside retrieved content.. Indirect prompt injection occurs when untrusted retrieved content is interpreted as instructions. The most effective design mitigation is to keep retrieved text in a clearly delimited data section and tell the model never to follow instructions found there. This changes the model's parsing context and makes injected commands less likely to be executed, while supporting output validation. Encryption and summarization do not address the trust boundary.

Answer

Ask the model to summarize each retrieved document before answering the user's question.

Answer

Store all documents in a vector database encrypted at rest with AES-256.

Answer

Increase the number of retrieved documents so the malicious content is diluted among many chunks.

Question 12

A team wants Claude to query their internal ticketing system through MCP. The operations are read-only lookups and the team wants the model to decide when a lookup is needed based on the conversation. Which MCP primitive is the appropriate choice?

Accepted Answer

MCP Tools exposed by the server.. Because Claude must decide on its own when to look up a ticket, the capability needs to be a callable action with a described schema, which is exactly what an MCP tool provides. Resources and prompts are not model-selected actions, and sampling reverses the request direction, so a tool is the correct primitive for this read-only lookup.

Answer

MCP Resources exposed by the server.

Answer

MCP Prompts exposed by the server.

Answer

MCP Sampling requests initiated by the server.

Question 13

A developer is creating an MCP server that exposes a tool to query a PostgreSQL database. The tool definition includes a parameter 'sql_query' that accepts a raw SQL string. During a security review, it was flagged that this design could allow SQL injection if Claude generates malicious SQL. Which change should the developer make to mitigate this risk?

Accepted Answer

Replace the raw SQL parameter with structured parameters such as table name, columns, and filters, and construct the SQL query internally using parameterized statements.. The best mitigation is to replace raw SQL with structured parameters and construct queries internally using parameterized statements. This eliminates the injection vector by design. Other options either rely on unreliable prompting, weak validation, or partial mitigation that does not address the core vulnerability. Structured parameters also improve usability and safety.

Answer

Limit the database user's permissions to read-only, so even if malicious SQL is executed, it cannot modify data.

Answer

Add a description to the tool instructing Claude to never generate malicious SQL.

Answer

Validate the 'sql_query' parameter by checking that it only contains SELECT statements and no semicolons.

Question 14

A fintech company deploys a Claude-powered agent that can call internal tools such as `get_transaction_history` and `initiate_transfer`. During a red-team exercise, an attacker crafts a user message that causes the agent to call `initiate_transfer` to an attacker-controlled account. Which TWO controls most directly mitigate this tool-abuse risk? (Choose two.)

Accepted Answer

Validate tool arguments against a server-side allowlist of permitted destination accounts and enforce authorization checks in the tool implementation.. Tool abuse is mitigated by enforcing trust boundaries outside the model. A human approval gate for high-impact actions prevents unauthorized execution, and server-side argument validation with authorization ensures the tool only acts on permitted inputs. Prompt instructions and temperature changes are not enforceable, and logging without prevention is only detective. Together, the two preventive controls block the attack even if the model is manipulated.

Answer

Add a system prompt that tells Claude to never call `initiate_transfer` unless the user explicitly asks for it.

Answer

Log all tool calls and review them weekly for suspicious patterns.

Answer

Increase the model's temperature to make its behavior less predictable to attackers.

Question 15

A developer is building an MCP client that connects to a remote MCP server over HTTP. After the initial handshake, the client needs to discover what prompts the server exposes so they can be surfaced in a slash-command menu. Which MCP capability should the client invoke?

Accepted Answer

prompts/list. Prompt templates are a distinct MCP primitive from tools and resources, and they are enumerated with the prompts/list request. Because the goal is to build a slash-command menu of reusable prompts, the client must call prompts/list after the initialize handshake, then optionally fetch a specific prompt with prompts/get when the user selects it.

Answer

resources/list

Answer

tools/list

Answer

initialize

Question 16

An agent built with the Anthropic SDK is processing a long conversation and begins to exceed the model's context window. The developer wants to preserve recent turns and key facts while reducing token usage. Which approach best matches how the SDK and Claude are designed to handle this?

Accepted Answer

Summarize older portions of the conversation into a compact message and prepend it, keeping the most recent turns verbatim.. Because the API is stateless, context management is the developer's responsibility. Summarizing older conversation segments while retaining recent turns verbatim balances token economy with continuity. This pattern is widely used in long-running agents to stay within the context window without losing the facts that matter.

Answer

Rely on the model's built-in long-term memory to automatically recall earlier turns without resending them.

Answer

Increase the temperature setting so the model can compress the conversation on its own.

Answer

Remove all prior messages and send only the latest user turn to avoid any risk of exceeding the limit.

Question 17

A developer is working in a large repository and wants Claude Code to understand the project structure and available tools before making changes. They run `claude` in the project root. Which action best ensures Claude Code has the necessary context for the session?

Accepted Answer

Run `/init` to generate a CLAUDE.md file that documents the project structure and common commands.. Using `/init` generates a CLAUDE.md file that persists across sessions and captures essential project details like build commands and architecture. This is the intended mechanism for providing durable context in a new codebase. Other options either do not exist, provide only temporary input, or do not enhance Claude Code's understanding of the project.

Answer

Manually paste the entire directory tree into the first prompt.

Answer

Run `claude --verbose` to enable detailed logging of all file accesses.

Answer

Execute `claude --reset` to clear previous session memory and start fresh.

Question 18

A developer is using the Claude API to generate creative writing pieces. The application often sends the same 2,000-token instruction set as a prefix to every request. Which feature can reduce the cost of processing this repeated prefix?

Accepted Answer

Prompt Caching. Prompt Caching is specifically designed to reduce costs when the same prefix is reused across multiple API calls. By caching the 2,000-token instruction set, subsequent calls avoid reprocessing those tokens at full input rates, directly lowering the cost. Other features like streaming, Batch API, or model selection do not target the redundant processing of a repeated prefix.

Answer

Batch API

Answer

Streaming responses

Answer

Using a smaller model

Question 19

A developer is optimizing a Claude-powered chatbot that handles 50,000 daily conversations. Each conversation starts with a 1,200-token system prompt that includes detailed instructions and examples. The system prompt is identical across all conversations. The developer wants to reduce input token costs. Which technique should be applied?

Accepted Answer

Enable prompt caching for the system prompt.. Prompt caching is designed for scenarios where a large, static prefix is reused across many requests. By caching the 1,200-token system prompt, the developer pays full price only once, and subsequent requests benefit from reduced input token costs. This directly targets the repetitive cost without sacrificing quality or altering the conversation flow.

Answer

Shorten the system prompt by removing examples.

Answer

Move the system prompt to the first user message.

Answer

Switch to a cheaper model for all conversations.

Question 20

A developer is using the Claude API to generate marketing copy for 50 different product lines. Each request includes a unique set of product details and a unique creative brief. The developer wants to minimize cost while ensuring high-quality output. Which model should they choose?

Accepted Answer

claude-3-sonnet-20240229. Marketing copy generation requires creativity and coherence but not the deep reasoning of Opus. Sonnet provides a balanced combination of quality and cost, making it the most suitable for this task. Haiku might sacrifice quality, while Opus would be overkill and more expensive. Claude 2.1 is outdated. Therefore, Sonnet is the optimal choice to minimize cost while ensuring high-quality output.

Answer

claude-3-opus-20240229

Answer

claude-2.1

Answer

claude-3-haiku-20240307