Learn the concepts
The ideas behind the exam, explained the same way the product teaches them — with a real practice question you can try right now.
- How Claude Code's agentic loop worksClaude Code runs on cycles of gathering context, acting, and verifying — not a fixed list of steps. See when delegating to subagents actually pays off.
- CLAUDE.md, hooks, and permissions: how Claude Code decides what it can doCLAUDE.md shapes Claude Code's behavior; permissions and hooks decide what it can actually execute. Understand the difference before configuring a project.
- Claude prompting techniques: XML tags, examples, and prefillA well-written prompt removes structural ambiguity. See how XML tags, examples, and prefill fix the most common mistakes when prompting Claude.
- What MCP is and how to write tools Claude picks correctlyMCP standardizes how Claude connects to external data and actions. See how to name and describe tools so it picks the right one, even among dozens.
- How Claude's prompt caching works — and how not to overpayPrompt caching cuts real cost, but only if the prefix is byte-for-byte identical. See where to put the breakpoint and how to handle 429, 529, and 400 errors.