Claude Code, honestly explained: the 10 slash commands you do need
What's making the rounds in my feed
Yesterday a LinkedIn post scrolls up at me: "The 5 secret Claude codes that make your productivity explode." Screenshots, big claims, 600 likes.
I open the thing and read: /ghost, /punch, X10, HARDMODE, /skeptic. Sounds like insider tips from a developer chat.
Except: half of them aren't commands at all.
They're plain words. You type them into your prompt, and Claude acts as if it understands them. Because Claude understands language. Not because some secret mechanism is running.
It's roughly like someone letting you in on this: "If you say PLEASE MAKE IT TASTY to the waiter, he brings you better food." Could happen. But it's not code. It's politeness.
What a real slash command is
Claude Code is Anthropic's terminal tool. You install it, open it in the terminal, and work with Claude right inside your project folder. For people who build websites, automate content workflows, write code.
It has real commands. They start with / and they do something for real. Not because Claude is nice. Because the tool executes them.
Here are the 10 I use every day. No tricks. No myths. Just tools that work.
- /clear
Empties the chat. You spent two hours building a website, now you want to do something else. /clear cleans up. Claude forgets the old stuff and starts fresh.
Why does that help? The fuller the chat, the more Claude "thinks" back. It gets slower and makes more mistakes. /clear is like a fresh sheet of paper on your desk.
- /compact
For when you want to stay on topic but the chat gets too long. Claude summarizes everything said so far and keeps working with the short version. The context stays, the dead weight goes.
The difference from /clear: you don't forget anything. You only compress.
- /model
Switches the model mid-session. Need fast answers for brainstorming? /model sonnet. Now a tricky problem that needs depth? /model opus.
Say you have a screwdriver and a cordless drill in your toolbox. For one screw, the first one does the job. For 100 screws, you want the other. Same principle.
- /init
Creates a CLAUDE.md in your project folder. That's the manual for Claude: where things are, what the project is called, which rules apply.
Once it's written, Claude reads it every time it starts. You don't have to explain every day that your website lives in /website and that you work in German.
- /memory
Opens Claude's long-term memory. This is where you enter what should apply across projects: "I work in German." "I don't like emojis." "My audience is freelancers and solo business owners, not techies."
Enter it once, and it stays on. No repeating yourself.
- /agents
Sets up subagents for you. These are small specialist Claudes for recurring tasks. One for research, one for SEO checks, one for writing on LinkedIn.
Each one has its own rules, its own voice, its own job. Like a team you don't have to pay.
- /hooks
Automation in the background. Examples: "Run the linter after every save." "Take a screenshot before every commit." "Make a backup first when closing."
Claude takes care of things you'd otherwise forget. Like a butler who locks the door before you walk to the car.
- /resume
Picks up an old conversation. Yesterday you worked on a feature, today you want to keep going. /resume, and Claude knows again where you left off.
No "remind me again what we did" loop. You just keep going.
- /cost
Shows you what the session has cost so far. Uncomfortable but honest: Claude, Opus in particular, costs money. If you don't know your costs, you're in for a surprise at the end of the month.
I check it several times a day. Not out of stinginess. Out of awareness.
- /review
Has Claude check your code. Just built something? /review, and it goes through it, finds bugs, suggests improvements. Like a senior colleague taking a quick look.
It doesn't replace a proper code review. But it catches 80% of careless mistakes before anyone else sees them.
What all this means
Most "AI productivity tips" in your feed are prompt cosmetics. Sometimes they work, sometimes they don't. It's language commenting on other language. Nothing more.
Real slash commands are different. They do something for real. /clear cleans up the context. /compact saves you tokens. /cost shows you the bill. No hocus-pocus.
If you don't use Claude Code yet: it's for anyone who regularly works on projects at a computer. Websites, content, automations, code. Installation takes five minutes.
If you already use it: setting up /init, /memory and /agents properly once gives you the biggest lever. Everything else builds on that.
That's it. No secret codes. No myths. Only tools that work once you know their names.
And an honest question: how many "prompt hacks" from your feed work for real? Or are you scrolling through a mythology right now?
FAQ
What is a slash command in Claude Code?
A slash command is a command that starts with a forward slash and that Claude Code executes as a program function. So it triggers a real technical action in the tool, for example clearing the chat or showing the costs. That sets it apart from normal prompt words, which Claude only interprets as text.
Why don't things like /ghost or X10 work for real?
These terms aren't real commands of the tool. They're normal words that end up in the prompt. Claude reacts to them because it understands language and tries to give a sensible answer, not because a fixed function sits behind them. The result is unreliable and depends on how the words happen to get interpreted.
What is Claude Code, anyway?
Claude Code is a terminal tool from Anthropic that lets you work with Claude right inside your project folder. It's aimed at people who build websites, automate content workflows or write code. You install it through the terminal, and that takes a few minutes.
What's the difference between /clear and /compact?
/clear deletes the entire chat history and starts over from scratch, so Claude forgets everything that came before. /compact, on the other hand, summarizes the chat so far and keeps working with that short version, so the context is preserved. The difference comes down to whether you start over or only cut dead weight.
What do I need a CLAUDE.md file for?
A CLAUDE.md is a kind of manual that Claude reads automatically every time it starts. It holds info like project structure, where files are stored, or working language. That way you don't have to repeat these details in every new session.