Lesson 3 of 4 · Claude Certified Architect — Foundations
The Six Exam Scenarios
Let's examine each scenario in detail so you know what to expect.
Scenario 1: Customer Support Agent
A company deploys a Claude-powered support agent that handles billing inquiries, processes refunds, troubleshoots technical issues, and escalates to human agents when needed.
This scenario tests your ability to design agentic loops with safety constraints. Expect questions about:
- Prerequisite gates (verify before refund)
- Escalation triggers (ambiguous identity, angry customer)
- Structured handoff to human agents
- Workflow enforcement via hooks vs. prompts
Scenario 2: Code Generation Pipeline
A development team builds a system that generates, reviews, and tests code across multiple programming languages using Claude.
Tests tool design and structured output:
- Tool interface design for code operations
- Error categorization (syntax vs. semantic vs. runtime)
- Multi-pass review (generate, then independently review)
- Structured error responses with actionable feedback
Scenario 3: Multi-Agent Research System
An organization builds a research system where a coordinator agent delegates tasks to specialized subagents for data gathering, analysis, and report generation.
Tests orchestration and context passing:
- Hub-and-spoke architecture
- Subagent isolation (no shared memory)
- Explicit context passing with structured metadata
- Iterative refinement loops
Scenario 4: Developer Productivity Tool
A team configures Claude Code for a large monorepo with multiple services, each with different conventions and requirements.
Tests Claude Code configuration:
- CLAUDE.md hierarchy (user/project/directory)
- Custom commands and skills
- Plan mode vs. direct execution
- Path-specific rules for different services
Scenario 5: CI/CD Quality Pipeline
A company integrates Claude into their CI/CD pipeline for automated code review, test generation, and release notes.
Tests automation and prompt engineering:
- Non-interactive mode (
-pflag,--output-format json) - Session context isolation
- Explicit review criteria
- Batch processing patterns
Scenario 6: Structured Data Extraction
A company processes thousands of documents (invoices, contracts, medical records) to extract structured data using Claude.
Tests reliability and context management:
- Schema design with nullable fields
- Confidence scoring and provenance tracking
- Error propagation vs. silent suppression
- Batch processing with the Message Batches API
Use ← → to navigate, Space to mark complete