Here’s a concrete hands-on roadmap to become a Claude Code power user (bookmark and use it over the next few days, progress saved client-side):
- Level 0: Setup Claude Code in your terminal. Hit session token limits.
- Level 1: Use CC to understand an unfamiliar codebase (can be an existing project at work or an open-source repo).
- Level 2: Go back to a previous checkpoint (hint: press Esc twice or use /rewind).
- Level 3: Try out 10 different slash commands (/bug, /compact, /clear, /help, /fork, /btw, /context, /export).
- Level 4: Explore and personalize /config to your liking.
- Level 5: Try voice coding using Wispr Flow (Fn) or the built-in voice mode (Space).
- Level 6: Save 3 useful memories (your preferred language, coding style, or workflow) and verify Claude uses them in a fresh session (hint: just like ! triggers bash mode, # can be used to add a memory).
- Level 7: CC navigates your browser to test stuff!
- Level 8: Get notified when it’s waiting for your input (hint: add a custom hook!).
- Level 9: Fire up 2 claude agents on the same codebase to work on independent features.
- Level 10: Integrate context7 MCP server.
- Level 11: Have Claude generate Mermaid architecture diagrams directly in markdown artifacts (hint: find and install a skill that does that).
- Level 12: Build your first reusable Claude skill (custom slash command).
- Level 13: Create a dedicated subagent for a specialized task (e.g. refactoring engineer or devil’s advocate).
- Level 14: Run multiple subagents in parallel to tackle different parts of a problem.
- Level 15: Ship a production-ready feature from idea → design → implementation → testing → documentation → PR using Claude Code as your primary engineering partner.
Use this list to calibrate yourself and see where you stand right now. These are incremental exercises that will help you code faster with claude code without compromising on quality!
DM me your score, along with any question you have! Feel free to share it with your friends as well.