SLASH COMMANDS · SAYING WHAT YOU ACTUALLY WANT
Two kinds of things get typed into Claude Code:
/.The four to know on day one:
/help — list everything it can do. When in doubt, start here./clear — wipe the conversation and start fresh. Use it between unrelated tasks./init — look around the project and write a CLAUDE.md file describing it, so it starts every future session already knowing your project./model — choose which Claude model to use.The habit that matters most: be specific about the goal, not the keystrokes. Say what "done" looks like, and let it work out the steps.
Weak: “Fix my code.”
It doesn't know which code, what "fixed" means, or how it would know when it's done. It will guess — and guessing is where bad results come from.
Strong: “The login form lets you submit an empty password. Make the password required, show an error message under the field, and add a test for it.”
Look at what the strong version gives it: where the problem is, what is wrong, what done looks like, and how to prove it. None of that is technical jargon — it's just clarity.
The most useful sentence you can add to almost any request is: “…and show me what you changed before you change it.”
Match each slash command to what it does.
Which request will get you the best result?
If you're not sure what the tool can do, /help is the right place to start.
Your turn. Think of something on your computer you'd like changed — a file to rename, a list to sort, a page to fix. Write the request you'd give Claude Code. Say where, what, and what "done" looks like.
Write your response before comparing it with the example.
0/600