CCAR-F mock exam

Question 1, Agentic Architecture & OrchestrationYour agent loop exits whenever Claude's reply contains text, and users report tasks stopping halfway through. What should control the loop?

• AExit when the reply contains the word "done" or a similar completion phrase
• BRun a fixed five iterations on every task, then return whatever you have
• CContinue while stop_reason is "tool_use" and finish on "end_turn"
• DAsk the model at the end of each turn whether it believes it has finished

Question 2, Agentic Architecture & OrchestrationA coordinator spawns a synthesis subagent, which replies that it has no search results to work with, although the search subagent finished. What is wrong?

• AThe synthesis subagent needs the web search tool
• BSubagents have isolated context, so the findings must be included in the prompt
• CThe coordinator needs a larger context window
• DThe search subagent should have written to shared memory automatically

Question 3, Agentic Architecture & OrchestrationA refund must never run before identity is verified. The prompt says so, yet logs show it happened on 1 in 300 conversations. What do you change?

• AMove to a larger model that follows long instructions more reliably
• BBlock process_refund with a hook until verification has returned a customer ID
• CRepeat the rule in capitals at the top and the bottom of the system prompt
• DAdd more few-shot examples showing verification always happening first

Question 4, Tool Design & MCP IntegrationAn agent keeps picking search_orders when it should pick find_orders. Both are described as "looks up orders". What fixes this at the root?

• AAdd a line to the system prompt telling the agent to prefer find_orders in most cases
• BRemove one tool and add a retry whenever the result looks wrong
• CGive the agent several more specific order tools to choose between
• DRename and describe the tools so their purposes, inputs and boundaries differ

Question 5, Tool Design & MCP IntegrationAn MCP tool returns "Operation failed" for every error. The agent retries invalid input as often as it retries timeouts. What should the tool return?

• AAn empty result, so the agent treats the call as finished and moves on to the next step
• BA structured error with a category, a retryable flag and a readable description
• CAn HTTP 500 status with no message body, so the client library raises its own exception
• DA longer free-text message that apologizes for the failure and asks the agent to try again

Question 6, Claude Code Configuration & WorkflowsA new hire says Claude Code ignores your team's conventions, though it works for everyone else. The conventions are in the tech lead's ~/.claude/CLAUDE.md. What is the fix?

• AMove the conventions into the project CLAUDE.md under version control
• BAsk the new hire to copy the tech lead's file into their own home directory
• CAdd the rules to each developer's shell profile so they load at startup
• DRun /compact at the start of each session so the rules stay in context

Question 7, Claude Code Configuration & WorkflowsClaude Code hangs in your CI job waiting for input. Which change fixes it and also gives you machine-readable findings?

• AAdd a CLAUDE.md rule telling it never to ask questions and to print JSON
• BPipe "yes" into the process so every prompt is accepted automatically
• CRun it with -p and use --output-format json with --json-schema
• DIncrease the CI job timeout to an hour and parse the terminal output

Question 8, Prompt Engineering & Structured OutputYour extraction schema marks po_number as required. On documents with no PO number, the model invents one. What is the best fix?

• AForce tool_choice to the extraction tool so the schema is always applied
• BMake the field nullable so the model can report it as absent
• CLower the temperature to zero so the model stops guessing at values
• DAdd a validation retry loop that rejects any PO number it cannot confirm

Question 9, Prompt Engineering & Structured OutputWhich workload should not use the Message Batches API?

• AA nightly test generation run that developers review the next morning
• BA weekly compliance audit of every contract in the repository
• CAn overnight analytics report that is read at the start of the day
• DA pre-merge check that blocks the developer until it returns

Question 10, Context Management & ReliabilityA customer says "I want to talk to a person" about a simple address change the agent could finish in seconds. What should the agent do?

• ACheck the sentiment score first, and hand off only if it is clearly negative
• BHand off to a human straight away, with a structured summary
• CComplete the change first since it takes seconds, then offer a human
• DExplain that the issue is simple and carry on unless they ask twice