Skip to main content
Photo of DeepakNess DeepakNess

Claude Code tips from the creator

Unproofread notes

Claude Code creator Boris shared some tips and best practices to using Claude Code, and it seems actually helpful. I am yet to get a Claude Code subscription but noting this down here for the future:

  1. you can do more things in parallel by using multiple git worktrees at once
  2. always start in plan mode, unless you're doing something very simple, and you can even ask another Claude instance to review and modify the plan and then only implement
  3. keep your CLAUDE.md file updated, end your prompts with "update your CLAUDE.md file so you do not make that mistake again"
  4. for each repetitive tasks, create a skill and can re-use across projects, can setup /slash commands to quickly refer to skills
  5. get Claude to fix the bug by itself: enable Slack MCP, paste a Slack bug thread into Claude and just say "fix"
  6. use prompts like:
    • grill me on these changes and don't make a PR until I pass your test
    • prove to me this works, etc.
  7. use /statusline to customize your status bar to always show context usage and current git branch
  8. add "use subagent" to your prompt where you want Claude to throw more compute at the problem
  9. if you also want to learn while coding, enable the "explanatory" or "learning" output style in /config
  10. ask Claude to explain the code by creating visual HTML presentation
  11. ask Claude to draw ASCII diagrams to understand the codebase in a better way

I can see that some of these are really useful, and not just for Claude Code but for other AI-assisted coding tools as well.

Comment via email