Free Practice Test

Your coordinator spawns subagents to research different aspects of a topic. You want the subagents to run concurrently rather than sequentially. How should you achieve this?

Your agent needs to handle mid-process escalation to human agents. When escalating a complex return case, what information should the handoff include?

Your agentic code generation pipeline reads a specification, generates code, runs tests, and fixes any failures. The loop currently runs until all tests pass or a maximum of 15 iterations is reached. In production, most tasks complete in 3-4 iterations, but some enter a cycle where the same test fails repeatedly with the same fix attempt. What should you add?

Your support agent's loop sometimes stops before the refund tool runs, and sometimes keeps going after the customer is answered. Which TWO control-flow rules are correct?

You want to force your document extraction agent to always call the extract_metadata tool first before any enrichment tools. Which tool_choice configuration achieves this?

Your team uses a shared GitHub MCP server for code review workflows and a personal experimental Playwright MCP server for testing. Where should each be configured?

Your CI pipeline needs to search the codebase for all TODO comments, read the surrounding context, and generate a prioritized task list. Which combination of built-in tools is most appropriate?

The agent keeps sending web results to analyze_document and PDFs to analyze_content. Both descriptions are one line and nearly identical. Which TWO fixes should come first?

You want to create a custom /review slash command available to every developer who clones your repository. Where should you create the command file?

Your CI pipeline script runs claude "Review this PR for security issues" but the job hangs indefinitely. Logs show Claude Code is waiting for interactive input. What's the fix?

A developer asks Claude Code to "fix the login bug" — a single, well-understood issue in one file where a null check is missing. The fix is 3 lines of code. What execution mode should Claude Code use?

A new teammate's Claude Code ignores your conventions, and the instructions file you rely on has grown to 900 lines. Which TWO changes fix this?

Your CI pipeline generates code and then reviews it in the same Claude Code session. The review step consistently approves the generated code, even when a human reviewer later finds obvious issues. What's the most likely cause?

You want to guarantee that your extraction agent calls a tool (not just returns text) but don't care which specific extraction tool it uses — the document type determines which tool is appropriate. What tool_choice setting should you use?

Your overnight batch extraction pipeline processes 1,000 contracts. After processing, you find that 50 failed: 30 had OCR quality issues, 15 exceeded token limits, and 5 had schema validation failures. How should you handle the failures?

Developers ignore your review bot: its "inaccurate comment" findings are mostly false positives, and its output format changes from run to run. Which TWO changes help most?

The web search subagent times out while researching a topic. How should this failure flow back to the coordinator?

A customer says "I want to talk to a real person." Your agent has already identified the issue (a standard return within policy) and could resolve it in one step. What should the agent do?

Your automated extraction pipeline achieves 97% overall accuracy. However, a recent audit reveals that accuracy on tax documents is only 78% while other document types exceed 99%. What should you do?

In long support sessions, summaries lose order numbers and refund amounts, each order lookup adds 40+ fields, and findings buried in the middle of long inputs get missed. Which THREE fixes apply?