Beginner-to-Competent in Claude Code

Intro

I’ve been noticing an increase in usage of Claude Code, likely after word spread around about Opus 4.5 performance being maximized when used in Claude Code (I’ve written about this in a previous blog post: Claude Code vs Cursor — My Take on this Debate).

Basics

• Start Claude Code from a terminal window with claude.
• Use claude --dangerously-skip-permissions to avoid having to grant permissions for every system command. There have been instances where people have gotten their home directories deleted by Claude Code, but they have been rare.
• Use subagents to do work from a terminal window with concurrency. You can just create subagents with natural language (“Use parallel subagents to do this task”).
• Install Skills to enhance Claude Code’s capabilities. I’m using obra/superpowers to give Claude Code better plans, enabling me to think through software architecture decisions more.
• Expect flickering when creating large codebase changes with Claude Code.

My Pro Tips

• Initialize a CLAUDE.md in your project with /init (Claude Code automatically recommends this when starting it in a new project folder).
• Run Claude Code in multiple terminal windows to work on tasks concurrently.
• You can use git worktrees to isolate clean git states while working on multiple tasks in parallel.
• When Opus 4.5 can’t solve a problem:
  • Activate thinking mode variants using natural language: “think”, “think hard”, “think harder”, “ultrathink”.
  • Prepend your prompts with: “You are a ”.
• Customize Claude Code with /config.

Claude Code is a tool that encourages the use of git and parallel tasks. With basic knowledge of Claude Code and utilizing what I wrote in Pro Tips, you are likely to see greatly increased productivity.