Exam Scenarios

The exam draws 4 of these 6 scenarios at random; each frames a set of questions. Descriptions follow Exam Guide v1.0 §5.

Scenario 1

Customer Support Resolution Agent

D1D2D5

You are building a customer support resolution agent with the Claude Agent SDK. It handles high-ambiguity requests such as returns, billing disputes and account issues, using custom MCP tools (get_customer, lookup_order, process_refund, escalate_to_human). The target is 80%+ first-contact resolution while knowing when to escalate.

What to focus on

  • Agentic loop and tool sequencing (verify the customer before refunds)
  • Programmatic prerequisites/hooks vs prompt instructions
  • Escalation triggers: explicit human request, policy gap, no progress
  • Structured error responses and persistent case facts
▶ Watch the video for this topic (D1)▶ Watch the video for this topic (D2)

Scenario 2

Code Generation with Claude Code

D3D5

You are using Claude Code to accelerate software development: code generation, refactoring, debugging and documentation. You integrate it into your workflow with custom slash commands, CLAUDE.md configuration, and understand when to use plan mode vs direct execution.

What to focus on

  • CLAUDE.md hierarchy, @import and .claude/rules/ with paths: frontmatter
  • Skills and slash commands
  • Plan mode vs direct execution
  • Managing long sessions (/compact, subagents)

Scenario 3

Multi-Agent Research System

D1D2D5

You are building a multi-agent research system with the Claude Agent SDK. A coordinator agent delegates to specialized subagents: one searches the web, one analyzes documents, one synthesizes findings, and one generates reports. The system produces comprehensive, cited reports.

What to focus on

  • Coordinator–subagent decomposition and explicit context passing
  • Parallel Task calls; allowedTools must include "Task"
  • Error propagation with structured metadata
  • Claim-source mapping and provenance
▶ Watch the video for this topic (D1)▶ Watch the video for this topic (D2)

Scenario 4

Developer Productivity with Claude

D2D3D1

You are building developer productivity tools with the Claude Agent SDK. The agent helps engineers explore unfamiliar codebases, understand legacy systems, generate boilerplate and automate repetitive tasks, using built-in tools (Read, Write, Bash, Grep, Glob) and integrating MCP servers.

What to focus on

  • Choosing built-in tools (Grep for content, Glob for paths)
  • MCP server configuration (.mcp.json, ${VAR} expansion)
  • Incremental codebase exploration and scratchpad files
  • Tool scoping per agent
▶ Watch the video for this topic (D2)▶ Watch the video for this topic (D1)

Scenario 5

Claude Code for Continuous Integration

D3D4

You are integrating Claude Code into a CI/CD pipeline. The system runs automated code reviews, generates test cases and gives pull-request feedback. Prompts must give actionable feedback and minimize false positives.

What to focus on

  • -p (print/headless) mode, --output-format json and --json-schema
  • Explicit review criteria to cut false positives
  • Independent review instance rather than self-review
  • Batch API vs synchronous calls for CI latency

Scenario 6

Structured Data Extraction

D4D5

You are building a structured data extraction system with Claude. It extracts information from unstructured documents, validates output with JSON schemas and keeps high accuracy. It must handle edge cases gracefully and integrate with downstream systems.

What to focus on

  • tool_use with a JSON schema and forced tool_choice
  • Optional/nullable fields to avoid fabricated values
  • Validation-retry loops with specific error feedback
  • Field-level confidence calibrated on labeled sets; stratified sampling