Scenario-focused practice built on two scenarios from the official exam guide.
Scenario 1 — Customer Support Resolution Agent (Q1–12): a Claude Agent SDK support agent wired to
custom MCP tools (get_customer, lookup_order, process_refund,
escalate_to_human), targeting 80%+ first-contact resolution while knowing when to escalate.
Scenario 2 — Code Generation with Claude Code (Q13–25): using Claude Code for code generation, refactoring, debugging, and documentation — custom slash commands, CLAUDE.md configuration, and plan mode vs direct execution.
Scenario 3 — Multi-Agent Research System (Q26–38): a coordinator agent delegating to specialized subagents (web search, document analysis, synthesis, report generation) to produce comprehensive, cited reports.
Scenario 4 — Developer Productivity with Claude (Q39–51): using Claude Code to explore unfamiliar codebases with built-in tools (Read/Write/Edit/Bash/Grep/Glob), integrate MCP servers, and configure project context, permissions, and sessions for everyday development.
Scenario 5 — Claude Code for Continuous Integration (Q52–64): automated PR reviews, test generation, and CI feedback — explicit criteria and structured output to minimize false positives and gate the build deterministically.
Scenario 6 — Structured Data Extraction (Q65–77): extracting fields from unstructured documents, constraining output with JSON schemas, and handling edge cases — validation/retry loops, confidence, provenance, and honest handling of missing or out-of-scope data.
All items are all-new (not Mock #1/#2) and modelled on the official sample-question style — pick the option a strong architect would choose, not just a plausible one. Pass flagged at 72% as a proxy for the real 720/1000; the "right" answer is usually the proportionate fix, not the most elaborate one.