Lesson 43: Checkpoint — Advanced Patterns Review
Day 43 of 50 · ~15 min assessment · Phase 5: Advanced Patterns
Your Goal
Test your understanding of Phase 5 concepts by answering 16 questions across all five phases. This checkpoint validates that you're ready for the final phase of the curriculum.
Scoring: 16 total questions. Aim for 14/16 or higher (88%) to move forward confidently.
Phase 1 Foundations (2 questions)
Question 1
You're in a Claude Code session reviewing a codebase. After 30 minutes of conversation, you notice your context window is nearly full. What's the best next action?
A) Ask Claude to summarize everything and keep going
B) Create a new session so you have fresh context
C) Use /compact to compress your current context
D) Both B and C are good options depending on the task
Question 2
In Claude Code, what's the fundamental difference between reading code and editing code?
A) Reading costs tokens; editing costs money
B) Reading is passive; editing requires human approval in interactive mode
C) There is no difference; they both use the same tools
D) Editing is only possible in headless mode
Phase 2 Intermediate (2 questions)
Question 3
You're setting up Claude Code for a production codebase. In your CLAUDE.md, you want to prevent Claude from deleting critical files. What's the best approach?
A) List all the files Claude shouldn't delete
B) Use directory-level permissions to forbid edits in critical directories
C) Trust Claude to not delete important files
D) Don't use Claude Code on production (too risky)
Question 4
What's the purpose of a hook in Claude Code?
A) To integrate with GitHub webhooks
B) To trigger Claude Code automatically when files change
C) To prevent Claude from accessing certain files
D) To cache context between sessions
Phase 3 Setup (2 questions)
Question 5
When you use MCP (Model Context Protocol), what are you actually doing?
A) Creating a new context window for Claude
B) Connecting Claude Code to external tools and data sources
C) Modifying Claude's underlying model
D) Compressing context to save tokens
Question 6
You want to create a reusable workflow (something you'll use across multiple projects). Which Claude Code feature is designed for this?
A) Skills
B) Hooks
C) Subagents
D) Channels
Phase 4 Execution (3 questions)
Question 7
You're debugging a complex bug where Claude Code proposed a fix, but tests are still failing. What's the best debugging approach?
A) Start over with a new session
B) Ask Claude to read the test output and propose a new fix
C) Use extended thinking to reason through the problem more carefully
D) Give up and fix it manually
Question 8
You're running tests in Claude Code and they're failing intermittently. What's the most likely cause?
A) Claude Code has a bug
B) Your codebase is flaky
C) Tests are timing-dependent or have race conditions
D) Claude Code doesn't run tests properly
Question 9
In a checkpointing workflow, when should you save a checkpoint?
A) After every single change
B) After completing a major milestone or feature
C) Only when you encounter an error
D) Never; checkpointing is overkill
Phase 5 Advanced Patterns (7 questions)
Question 10
You need to do three independent tasks: security review, test coverage analysis, and documentation generation. All are CPU-intensive and could take 10 minutes each. Should you use subagents?
A) No, just do them sequentially in one session
B) Yes, spawn three subagents to run them in parallel
C) Yes, but only if each one is a complete, autonomous task
D) Never use subagents; they're too complex
Question 11
You set up headless mode for Claude Code to run PR reviews in GitHub Actions. The first time it runs, it makes a destructive change (deletes a file it shouldn't). What went wrong?
A) Headless mode is inherently unsafe
B) You didn't use --print to preview before running
C) You didn't configure permissions properly (CLAUDE.md)
D) GitHub Actions doesn't support Claude Code
Question 12
You're optimizing costs and notice that reading a 5,000-line file costs the same whether you read the whole thing or just the 50 lines you need. What's the best strategy?
A) Always read whole files; it doesn't matter
B) Read selectively and save tokens
C) Use a cheaper model to offset costs
D) Use subagents to avoid reading at all
Question 13
You're working on a 200K-line monorepo and Claude is struggling to understand the architecture. What's the most important first step?
A) Read everything in the codebase
B) Create a detailed CLAUDE.md architecture map
C) Use subagents to explore different modules
D) Give up; the codebase is too large
Question 14
What's the key difference between Claude Code's "harmless if untrusted" design and "trustworthy" design?
A) They're the same thing
B) Harmless if untrusted means architecturally constrained; trustworthy means relying on the model's behavior
C) One applies to headless mode; the other to interactive
D) Trustworthy is better; harmless is for paranoid teams
Question 15
You want to control Claude Code from your phone while in a meeting. What's the right tool for this?
A) Subagents
B) Headless mode
C) Channels (Telegram, Discord, iMessage)
D) Cloud providers
Question 16
Your company uses AWS exclusively, and there's a requirement that all code stays within AWS boundaries. Which cloud provider integration makes the most sense?
A) Direct Anthropic API
B) AWS Bedrock
C) Google Vertex AI
D) Microsoft Azure Foundry
Scoring
Count your correct answers:
- 14-16: Excellent! You understand advanced patterns deeply. Ready for Phase 6.
- 12-13: Good! You have solid understanding but review the weaker areas.
- 10-11: Fair. Review Phase 5 lessons before moving on.
- Below 10: Consider revisiting the Phase 5 lessons before Lesson 44.
What's Next?
If you scored 14+, you're ready for the final lesson: Troubleshooting.
This lesson teaches you how to diagnose and fix common problems with Claude Code — subagent failures, runaway costs, infinite loops, and more.