Your terminal is the new workspace.
I know that sounds dramatic. But once you start using Claude Code, you'll realize most of your time is wasted on clicking, switching tabs, and re-explaining context. These 10 commands fix that.
This is the cheatsheet I wish I had when I started. No fluff, just the commands that actually matter for marketers working with Claude Code every day.
TL;DR
--dangerously-skip-permissions— Go full autopilot, no permission prompts/clear— Reset the conversation, keep your instructions/btw— Side question without breaking Claude's flow--resume— Browse and reopen past sessionsCtrl + S— Stash your half-typed prompt, pop it back later@filename— Autocomplete file paths in your prompt-c— Continue your last conversationShift + Tab— Cycle permission modes without restarting! command— Run shell commands from the Claude prompt/cost— See token usage and cost for your session
1. Go full autopilot
$ claude --dangerously-skip-permissions
Normally Claude asks permission before reading files, running commands, or editing code. This flag skips all of that.
Claude just... does the work. No interruptions, no "are you sure?" popups. You sit back and watch.
When to use it: Personal projects where you trust Claude to do its thing. Not recommended for production code you don't review. I use this for brainstorming sessions, content creation, and exploring data where I want zero friction.
Tired of typing that monster command every time? Add this line to your
~/.zshrc(or~/.bashrc):
alias yolo="claude --dangerously-skip-permissions"Now just type
yoloand you're in. Runsource ~/.zshrconce to activate it. You're welcome.
2. Wipe the slate clean
/clear
Clears the entire conversation from Claude's memory. Fresh start, zero baggage. Your CLAUDE.md instructions stay loaded.
Think of it as closing all your browser tabs and starting over. Sometimes that's exactly what you need.
Pro tip: Use this when Claude starts giving weird answers or goes in circles. A clean context often fixes it instantly. I probably use this 5-10 times a day.
3. Ask without interrupting
/btw what file was that again?
Quick side question while Claude is working. It can see your conversation but won't derail whatever task is running.
The answer is ephemeral. It never enters the conversation history. Zero context cost.
Perfect for: "Wait, what was the name of that campaign?" or "Which folder did we save that to?" without breaking Claude's flow. You get your answer, Claude keeps working. Everybody wins.
4. Browse and resume any session
$ claude --resume
Opens an interactive list of all your past conversations. Scroll through, pick the one you need, and you're right back in it.
Way more useful than -c (which just reopens the last one) when you have multiple projects going. You choose which session to reopen.
Pro tip: Name your sessions with claude -n "q2-campaign" so you can spot them instantly in the list later. Future you will thank present you.
5. Stash your message
Ctrl + S
Halfway through typing a prompt? Hit Ctrl+S to stash it. Ask Claude something else first. Hit Ctrl+S again to pop your original message back.
Like git stash, but for your prompt. Your half-written message is safe while you handle something quick.
Example: You're writing a long brief, realize you need to check a file first. Stash it, ask your question, pop it back. No retyping. No losing your train of thought.
6. Point Claude to files instantly
> read @campaigns/q2-launch.md
Type @ and start typing a filename. Claude autocompletes the path for you. No more copy-pasting file locations.
Works anywhere in your prompt. Attach context, reference docs, or point to specific files you want Claude to work with.
Example: "Rewrite the headlines in @ads/search-campaigns.csv using our brand voice from @docs/brand-guide.md." Two files referenced, zero paths memorized.
7. Pick up where you left off
$ claude -c
Closed the terminal? Went to lunch? claude -c continues your last conversation with full context intact.
No need to re-explain what you were working on. It's like reopening a paused Netflix episode. Everything is exactly where you left it.
Also try: claude -r "campaign-audit" to resume a specific named session. Pair this with claude -n "my-task" to name sessions when you start them.
8. Switch permission modes on the fly
Shift + Tab
Cycle between permission modes without restarting. Go from "ask me everything" to "plan mode" to "auto mode" with one shortcut.
Start careful, then loosen the reins once you trust what Claude is doing. No need to restart the session.
The modes:
- Default - asks before actions
- Plan - suggests, you approve
- Auto - just does it
I usually start in default mode, read through a few actions to make sure Claude understands what I want, then switch to auto. Best of both worlds.
9. Run terminal commands inline
! ls campaigns/
Start your message with ! to run a shell command directly from the Claude prompt. The output lands right in the conversation.
No need to switch terminal windows. Check files, run scripts, open folders. All without leaving your Claude session.
Handy examples:
! open .to open Finder! git statusto check changes! cat notes.mdto peek at a file
This one saves me more time than any other shortcut on this list. The context switch between terminals adds up fast.
10. Know what you're spending
/cost
Shows exactly how many tokens you've used in this session and what it cost. No more guessing if you're burning through your quota.
Essential for freelancers who bill AI usage to clients, or anyone who just wants to stay on budget.
Bonus commands:
/contextshows a visual grid of how full your context window is/usageshows your plan limits and rate status
The 80/20 of Claude Code
These 10 commands cover 80% of what you need to be fast in Claude Code as a marketer.
The other 20%? Just type /help and explore. Claude Code has a lot more depth, but you don't need all of it on day one.
Start with these. Get comfortable. Build the muscle memory. Then go deeper when you hit the edges.
The terminal is the new ads editor. The sooner you get comfortable in it, the faster everything else becomes.
About Alfred Simon
AI Systems Builder & Coach
I build custom AI systems for marketing teams — search term analysis, ad creation, competitor research, reporting — all automated. I write about context management, AI workflows, and the messy reality of building things with AI. No theory. No hype. Just what actually works after 30+ agents and a very healthy trash pile :D
Want to build something like this for your team? Let's talk.