Claude Code Tips: 30 Shortcuts, Slash Commands, and Automations

By Beau Johnson··8 min read

Quick answer

I cover 30 Claude Code tips: shortcuts like Shift+Tab, Esc and Ctrl+G, commands like /effort, /compact and /context, plus skills, MCP, subagents, hooks, /goal, /code-review, claude -p, /loop and /schedule so Claude can run whole jobs on its own.

In this video I go through 30 Claude Code tips in under 5 minutes, and I show every one in a real Claude Code terminal. It's for anyone who uses Claude Code like a chat box, where you type a prompt and wait, and wants to get to the point where Claude runs whole jobs on its own. Everything here was checked against the official docs for Claude Code 2.1.312 on November 6, 2026.

If you haven't installed Claude Code yet, start with the install lesson first. If you want more on the tool itself, my Claude Code page collects everything in one place.

What I cover in this video

  • 0:00 Why most people use Claude Code like a chat box, and the first tips for driving it faster (permission modes, plan mode, @ files, shell mode)
  • 0:46 Screenshots, voice, new lines and your text editor, then how to steer Claude while it works
  • 1:33 Rewind, stashing prompts, searching past prompts and background commands
  • 2:18 How hard it thinks and what it remembers: /effort, ultrathink, model switching, /clear, /compact, /context
  • 3:03 Teach it once: CLAUDE.md, skills, plugins, MCP and subagents
  • 3:50 Let it work without you: hooks, /goal, /code-review, claude -p, /loop and /schedule
  • 4:36 Bonus tip 31: how this video itself was made by Claude Code

Drive it faster and steer it mid-flight (tips 1 to 14)

The first group is about getting your instructions into Claude Code faster.

  • 1. Shift+Tab cycles your permission modes, so you pick how much Claude asks before it acts.
  • 2. Plan mode. Claude researches and writes a plan without editing your files until you approve, so you catch a bad idea early.
  • 3. @ points Claude at the exact file instead of making it search.
  • 4. ! puts you in shell mode. Run your tests and Claude reads the output.
  • 5. Ctrl+V pastes a screenshot from your clipboard. On a Mac it's Ctrl, not Cmd.
  • 6. /voice once (it needs a Claude.ai login), then hold Space and talk. Way faster than typing.
  • 7. Shift+Enter gives you a new line in most terminals, and Ctrl+J works in any terminal.
  • 8. Ctrl+G opens your prompt in your text editor for the long ones.

Then I cover steering Claude while it's working, so you don't have to sit and wait for a turn to finish.

  • 9. Esc stops it mid-turn and keeps what it's done so far, so you can redirect it the second it drifts.
  • 10. Type while it works. Hit Enter and it picks up your note after the current step.
  • 11. Esc twice on an empty prompt (or /rewind) rolls your code and chat back to an earlier prompt. Keep using git, though, because shell changes aren't tracked.
  • 12. Ctrl+S stashes a half-typed prompt. Ctrl+S on an empty prompt brings it back.
  • 13. Ctrl+R searches your past prompts so you stop retyping your best ones.
  • 14. Ctrl+B sends a long command, like a dev server, to the background so you keep working while it runs.

How hard it thinks and what it remembers (tips 15 to 20)

  • 15. /effort. Use low for drafts and high or xhigh for nasty bugs. The default on Opus 5.5 is medium.
  • 16. ultrathink in a prompt gives you one turn of deeper thinking, and your effort setting stays put.
  • 17. Option+P switches models without clearing your prompt. On a Mac, turn on "Use Option as Meta Key" in your terminal first.
  • 18. /clear between unrelated tasks. The old chat is still there in /resume.
  • 19. /compact summarizes the chat. Add a focus, like keep the API changes, and it holds on to that.
  • 20. /context shows you exactly what's filling your context window.

Teach it once so you never repeat yourself (tips 21 to 25)

  • 21. /init writes a starting CLAUDE.md. Keep it short: your commands and anything you've corrected twice.
  • 22. Skills. Anything you keep pasting into chat, turn into a skill and run it with /name. This video runs on skills.
  • 23. /plugin installs bundles of skills and hooks. Read a plugin before you install it, because it runs with your permissions.
  • 24. MCP connects Claude to your tools, like your issue tracker, so you stop copying and pasting. You add servers with claude mcp add and check them with /mcp.
  • 25. Subagents. Ask for a subagent on side research. It works in its own context and hands back a summary.

Let it work without you (tips 26 to 30)

The last five are where Claude Code stops being a chat box and starts running jobs. If you're getting into AI agents, this is the part to focus on.

  • 26. Hooks are for rules that can't break. Claude Code runs a hook every time, and the model doesn't get a vote.
  • 27. /goal sets a finish line, like all tests pass, and Claude keeps going until a check says it's done. Run it in auto mode if you're walking away.
  • 28. /code-review brings in a fresh reviewer that never saw your chat to check your branch before you ship.
  • 29. claude -p runs Claude once and exits, so you can drop it into your scripts.
  • 30. /loop and /schedule. /loop repeats a prompt while your session is open. /schedule runs it in the cloud with your laptop closed, on a fresh copy of your repo. That one is a research preview on paid plans.
claude -p

The bonus tip 31 is the video itself. Claude Code wrote it, ran my voice clone and edited it on a 4 AM job using skills and claude -p. That's tips 22 and 29 working together.

Key takeaways

  • Plan mode, Esc and rewind let you catch and fix a bad direction early instead of waiting for Claude to finish.
  • Use /effort and ultrathink to match how hard Claude thinks to the job, and /clear, /compact and /context to keep your context window clean.
  • Stop repeating yourself: a short CLAUDE.md, skills for anything you keep pasting, and MCP for your tools.
  • Read plugins before installing them, since they run with your permissions, and keep using git because rewind doesn't track shell changes.
  • Hooks, /goal, claude -p, /loop and /schedule are how Claude runs whole jobs without you sitting there.

FAQ

How do I stop Claude Code without losing its work?

Press Esc. It stops Claude mid-turn and keeps what it has done so far, so you can redirect it. To roll code and chat back to an earlier prompt, press Esc twice on an empty prompt or use /rewind, but keep using git because shell changes aren't tracked.

What is the difference between /effort and ultrathink in Claude Code?

/effort sets how hard Claude thinks going forward: low for drafts, high or xhigh for hard bugs, with medium as the default on Opus 5.5. Putting ultrathink in a prompt gives you one turn of deeper thinking and leaves your effort setting where it was.

Can Claude Code run tasks while my laptop is closed?

Yes, with /schedule, which runs a prompt in the cloud on a fresh copy of your repo. It's a research preview on paid plans. /loop repeats a prompt too, but only while your session is open.

How do I run Claude Code from a script?

Use claude -p. It runs Claude once and exits, so you can drop it into your scripts. This video was made on a 4 AM job using skills and claude -p.

Chapters

  1. 0:0030 tips in under 5 minutes
  2. 0:23Drive it faster (tips 1-8)
  3. 1:15Steer it mid-flight (tips 9-14)
  4. 2:01Night Crew
  5. 2:15How hard it thinks (tips 15-20)
  6. 3:01Teach it once (tips 21-25)
  7. 3:41Let it work without you (tips 26-30)
  8. 4:28Bonus tip 31

Transcript

Read the full transcript

0:00Most people use Claude code like a chat box. Type a prompt and wait. These 30 tips take you from that to Claude running whole jobs on its own in under 5 minutes. So, in this video, I'm breaking down the shortcuts, commands, and automation most people skip. If you haven't installed it, there's a step-by-step lesson in our school community linked down below. Let's jump right into it. Number one, shift tab cycles your permission modes. So, you pick how much Claude asks before it acts. Two, plan mode. Claude researches and writes a plan without editing your files, so you catch a bad idea early. Three, type the at sign to point it at the exact file instead of making it search. Four, an exclamation point puts you in shell mode, so you run your tests

0:46and Claude reads the output. Five, copy a screenshot to your clipboard, then controlV pastes it in. On a Mac, control, not command. Six, turn on /Voice once with a claude.AI login, then hold space and talk. Way faster than typing. Seven, shift enter gives you a new line in most terminals, and Ctrl J works in any terminal. Eight, Ctrl G opens your prompt in your text editor for the long ones. So now, let's steer it while it works. Nine. Escape stops it mid turn and it keeps what it's done so far so you can redirect it the second it drifts. 10. Don't wait. Type while it works. Hit enter and it picks up your note after the current step. 11. Escape

1:33twice on an empty prompt opens rewind to roll your code and chat back to an earlier prompt. Keep using Git, though. Shell changes aren't tracked. 12. Ctrl S stashes a half-typed prompt. Crl S on an empty prompt brings it back. 13. ControlR searches your past prompts so you stop retyping your best ones. 14. Control B sends a long command like a dev server to the background and you keep working while it runs. Every night, Night Crew researches what your market is talking about, writes the script, fact checkcks it, voices it in your voice, and animates it. In the morning, you approve it and it posts to YouTube, Instagram, Tik Tok, and Facebook. Nightcrew. Nightcrew.studio. studio. So, how hard does it think and what does it

2:18remember? 15/effort. Low for drafts, high or x high for nasty bugs. The default on Opus 5.5 is medium. 16. Ultraink in one prompt gives you one turn of deeper thinking and your setting stays put. 17. Option P switches models without clearing your prompt. On a Mac, turn on option as meta first. 18. slashclear between unrelated tasks. The old chat is still in /resume. 19 /compact summarizes the chat and if you add a focus like keep the API changes, it holds on to that. 20 /context shows you exactly what's eating your context window. Now let's teach it once so you

3:03never repeat yourself. 21 /init writes a starting claw.md. Keep it short, your commands and anything you've corrected twice. 22. Anything you keep pasting into chat, turn into a skill and run it with a slash. This video runs on skills. 23. Slash plugin installs bundles of skills and hooks. Read one before you install it. It runs with your permissions. 24. MCP connects Claude to your tools like your issue tracker so you stop copy pasting. 25. Ask for a sub agent on side research. It works in its own context and hands back a summary. And the last five, let it work without you. 26, hooks. So for rules that can't break, Claude code runs a hook every

3:50time and the model doesn't get a vote. 27 / goal sets a finish line, like all tests pass, and Claude keeps going until a check says it's done. Run it in auto mode if you're walking away. 28.code review. A fresh reviewer that never saw your chat checks your branch before you ship. 29. Claude-P runs Claude once and exits so you can drop it into your scripts. 30 / loop repeats a prompt while your session is open and /chedu runs it in the cloud with your laptop closed on a fresh copy of your repo. That one's a research preview on paid plans. So, bonus tip 31. Claude Code wrote this video, ran my voice clone, and edited it on a 4 a.m. job using

4:36Skills and Claude-P. Which of these 30 was new to you? Drop the number in the comments. And if one saved you some time, hit like. I'll keep making these quick ones. Have a blessed day.

Ready to start building with AI?

Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.

Join Shipping Skool

Keep reading