Brooke Wright
Brooke Wright · @wright_mode
FREE CHEAT SHEET

Twelve Claude Code codes that'll save you hours

The five codes from the reel plus seven bonus shortcuts I use every single day. Copy, paste, bookmark — the ones that actually matter.

12 codes 5 min read Built from real sessions Built by Brooke
Something went wrong. Please try again.

No spam. Unsubscribe anytime.

Wright Mode — Free Cheat Sheet

The Claude Code Codes Cheat Sheet

Twelve shortcuts that make you faster, cheaper, and less frustrated inside Claude Code. The ones I actually use every day — not the 80 you'll never touch.

⚡ 12 codes total 📋 Copy-paste ready 🎯 Built from real sessions ⏱ 5 min read

What's inside

Claude Code has 80+ slash commands and shortcuts. Most of them you'll never need. This cheat sheet is the twelve I reach for every single day — the ones that actually make a difference to how fast you move, how much you spend, and how often you want to throw your laptop at the wall.

"Codes" here means three things:

Which ones to memorise first

If you only learn five, make them the Big Five (next section). They'll cover 80% of your day. The Bonus Seven are for when you're ready to level up.

Works in the Claude Code CLI
Any plan — Pro, Max, Team
No extensions required
No fancy setup

Start here. These are the five I showed in the reel — the ones that changed how I work once I actually started using them. Drill them into muscle memory before you worry about the rest.

1

ultrathink — maximum reasoning

Add this word anywhere in your prompt to unlock Claude's deepest reasoning budget before it responds. It thinks longer, considers more angles, and catches things it'd otherwise miss. Use it for architecture decisions, nasty bugs, anything where being wrong costs you hours. Example: "ultrathink — review this database schema and flag any issues before I migrate."

2

/compact — free up context

Summarises the current conversation into a condensed version so you can keep going without losing the thread. Use it when sessions drag past an hour, or the moment you see "context getting large" in the corner. Saves you from starting over and losing everything Claude just learned about your project.

3

"be direct, don't glaze me" — kill the sycophancy

Add this as a prefix to any prompt where you need an honest answer. Claude's default is to be encouraging, which is lovely but useless when you're trying to figure out if your idea is actually bad. Use it every time you want real feedback — on code, on copy, on business decisions. I use it 10+ times a day.

4

/export — save your session

Exports your entire session as a markdown file you can save, share, or paste into Notion. Use it after you've built something you want to keep — a working script, a solid plan, a useful debugging thread. Future-you will thank you when you need to remember how you did something three weeks ago.

5

Esc Esc (or /rewind) — undo button

Rewinds the session to any earlier checkpoint — before Claude broke your file, before you gave it a bad instruction, before the whole thing went sideways. Use it when Claude does something destructive or heads down the wrong path. This is your "oh no" insurance. Way faster than trying to explain the fix.


The seven I didn't have room for in the reel. Add these to your toolkit once the Big Five feel natural. Every one of them saves you time, money, or frustration.

/clear

Start fresh, same session

Clears the chat history so you can start a new topic without restarting Claude Code entirely.

When: Switching topics — way faster than a full restart.
/model

Switch between models

Swap between Opus, Sonnet, and Haiku mid-session without leaving the chat.

When: Cheap tasks → Haiku. Hard stuff → Opus. Save tokens.
/cost

Check your token spend

Shows exactly how many tokens you've burned this session and roughly what it's cost you.

When: Before you panic at your monthly bill.
/plan

Plan before touching files

Claude plans the whole task out before writing a single line. Also accessible via Shift+Tab into Plan Mode.

When: Any multi-step task, especially risky edits.
# at start of message

Add a memory instantly

Starting your message with # adds that line to your CLAUDE.md memory file. Example: "# always use npm not yarn in this project".

When: You just corrected Claude and don't want to correct it again.
/resume

Pick up a past session

Shows a list of your previous sessions so you can jump back into one that's still relevant.

When: You closed the terminal and panicked.

Bonus #12 — Ctrl+C

The one keyboard shortcut in the bunch. Stops Claude mid-generation. Use it the second you see it heading the wrong direction — don't wait for it to finish the bad answer. Interrupting early saves tokens and your sanity, and you can redirect with a better prompt.


Screenshot this. Pin it. Bookmark it. The whole lot in one scannable grid — code, what it does, when to use it.

Code
What it does
When
ultrathink
Max reasoning budget
Hard problems
/compact
Summarises history, frees context
Long sessions
"be direct, don't glaze me"
Stops sycophancy
Real feedback
/export
Save session as markdown
Keep the good stuff
Esc Esc
Rewind to checkpoint
Claude broke it
/clear
Start fresh, same session
Topic switch
/model
Swap Opus / Sonnet / Haiku
Match model to task
/cost
Show token spend
Budget check
/plan
Plan before touching files
Risky edits
# at start
Save to CLAUDE.md memory
Rule for the project
/resume
Pick up a past session
Terminal closed
Ctrl+C
Stop mid-generation
Wrong direction

Pro tip: print this page

Hit Cmd+P (or Ctrl+P on Windows) — the page is print-optimised. The CTA bar, nav, and footer all drop away and you get a clean one-pager you can stick on the wall next to your desk.


Here's how I actually string these together in a real session. This is what a "build something" day looks like — from opening the terminal to closing it again without hating myself.

A typical Claude Code session

Session: "Build a Flask endpoint that processes a Fathom webhook"
1
Start with /plan — I let Claude outline the whole thing first before it touches a single file. Catches 90% of the "wait that's wrong" moments.
2
Prefix with ultrathink for the tricky bits — schema design, auth logic, anything where getting it wrong costs me hours.
3
Switch to /model Haiku for the boring stuff — writing tests, renaming variables, adding comments. Same job, fraction of the cost.
4
Hit Esc Esc when it goes sideways — and it will. Faster than explaining the mistake back to Claude.
5
Run /compact when the session's been going 45+ minutes. Keeps Claude sharp without losing what it's learned about the project.
6
/export before I close — so tomorrow-me has a paper trail.
7
Next day: /resume — pick up exactly where I left off. No "wait, what was I doing?"

The pattern

Plan first. Use the smartest model for the hardest parts. Drop to the cheapest model for the dumb parts. Rewind when wrong. Compact when long. Export when done. That's the whole game.


Three traps I see people fall into over and over. Avoid these and you'll be in the top 10% of Claude Code users inside a week.

Mistake 1: Running Opus for everything

Opus is expensive. If you're using it to rename variables or write a README, you're bleeding money. Switch to Haiku with /model for the grunt work. Save Opus for the hard stuff — architecture, debugging, anything where getting it right really matters.

Mistake 2: Never using /plan

Letting Claude start editing files without a plan is how you end up with a 40-minute session that rewrote your whole codebase to fix a two-line bug. /plan (or Shift+Tab into Plan Mode) forces Claude to think first, touch files second. Takes 30 seconds, saves hours.

Mistake 3: Letting bad responses finish

If you can see Claude is going in the wrong direction three sentences in, don't wait politely for it to finish. Hit Ctrl+C, give it a better prompt, move on. Every paragraph it writes after you've realised it's wrong is just money and time down the drain.

One more: not reading CLAUDE.md

Claude Code reads a file called CLAUDE.md at the start of every session — it's your project's rulebook. Drop your preferences in there (tech stack, spelling, testing conventions, whatever) and you stop repeating yourself. Use # at the start of any message to add a rule instantly. If you're correcting Claude more than twice on the same thing, you've got a CLAUDE.md problem, not a Claude problem.


Ready to actually use Claude Code?

These codes are the keyboard shortcuts. These resources are the whole keyboard.