The Claude Code cheat sheet: essential slash commands
This Claude Code cheat sheet starts with the slash commands I reach for every session. A slash command is a built-in shortcut you type starting with a forward slash, and it controls the session itself rather than asking the agent to build something. Learn these and you stop fighting the tool.
The table below describes what each command is for in plain language. I am deliberately not listing flags or exact arguments, because those shift between versions and I would rather you trust this than memorize something that breaks. When you want the full daily set with examples, see [the Claude Code slash commands you use every day](/blog/claude-code-slash-commands-you-use-every-day).
Essential Claude Code slash commands and what each one is for
| Command | What it does |
|---|---|
| /help | Lists the available commands and shows quick guidance, so it is the first thing to run when you forget a command. |
| /clear | Clears the current conversation context and starts fresh, which is what you want before an unrelated task. |
| /compact | Condenses the running conversation so a long session keeps its key context without filling up. |
| /init | Sets up the project so Claude Code understands the codebase, including creating the memory file it reads on startup. |
| /review | Asks Claude Code to review code, a useful checkpoint before you commit or hand work off. |
| /agents | Manages subagents - the specialised helpers a main agent can hand a focused job to. |
| /mcp | Shows and manages the connected MCP servers, the tools and data sources Claude Code can reach. |
| /config | Opens the settings for the session and your preferences, where you adjust how Claude Code behaves. |
| /cost | Reports the token usage and cost of the current session, so you can see what a run actually spent. |
| /model | Switches which Claude model the session uses, letting you trade speed for depth on demand. |
| /resume | Picks up a previous conversation so you can continue a session you left earlier. |
| /skills | Lists and manages the skills available to the agent - reusable packaged know-how it can load on demand. |
The four core concepts table: skills, subagents, MCP servers, hooks
Four concepts cover almost everything advanced you will do in Claude Code: skills, subagents, MCP servers, and hooks. Most people confuse them, then reach for the wrong one. This table gives each one a single row - what it is and when to use it - so you can pick fast.
Claude Code core concepts: what each is and when to use it
| Concept | What it is | When to use it |
|---|---|---|
| Skill | A reusable package of instructions and know-how the agent loads on demand. | When you repeat the same kind of task and want the agent to do it the same way every time. |
| Subagent | A focused helper agent the main agent hands a contained job to. | When a big task has a separable piece that deserves its own context and attention. |
| MCP server | A connector that gives Claude Code access to an outside tool or data source. | When the agent needs to reach something beyond your files, like a database, an API, or a service. |
| Hook | An automatic action the harness runs at a set point in the session. | When you want a behaviour to fire every time on its own, with no prompting from you. |
The trap is using a subagent when you needed an MCP server, or a skill when you needed a hook. For the full decision tree on which one to reach for, read [skills, subagents, MCP, and hooks: when to use each](/blog/claude-code-skills-subagents-mcp-hooks-when-to-use).
The CCC build flow: my everyday Claude Code loop
The CCC build flow is the five-step loop I run for almost every task, and it is desktop-app-first because that is where most members start. Memorize this and you have the only workflow you need on the cheat sheet. No commands required, just a clear pattern.
- Open Claude Code in the desktop app so you have a clean window and no terminal to wrangle.
- Point it at a real project folder so the agent has actual files to read for context.
- Give it a goal with a finish line: not 'clean this up' but 'rename these three functions and keep the tests passing'.
- Tell it how to verify: name the command to run, the build to pass, or the exact output you expect.
- Review what it did before you accept it - read the change, not just the summary.
Prompt-quality shortcuts that get better results
A Claude Code cheat sheet is not just commands - the biggest quality lever is how you write the prompt. Three habits do most of the work, and I run all three on every non-trivial ask. They cost ten extra seconds and save you a bad run.
- BLUF the goal: put the outcome you want in the first sentence, before any background. The agent weights the start of your message most.
- Scope the files: name the folder or files in play so the agent does not wander into code you did not mean to touch.
- Give it a check: hand over a test, a build, or an expected output so the agent can verify its own work instead of guessing.
Weak prompt versus scoped prompt
| Weak prompt | Scoped prompt |
|---|---|
| Make the login better | Add email validation to the login form in /auth and make the existing auth tests pass |
| Fix the bug | The cart total is wrong when a coupon applies; fix it in cart.ts and verify with the cart test |
| Clean up the code | Remove the unused imports in /components and confirm the build still passes |
Keep this Claude Code cheat sheet close
Screenshot the tables above and you have the working set: the slash commands, the four core concepts, the CCC build flow, and the prompt-quality shortcuts. That is the spine of using Claude Code well, and it stays true as the tool grows because it focuses on purpose, not on syntax that changes.
Everything past this cheat sheet is reps. Run the build flow on a real task today, scope your prompt, and give the agent a way to check itself. The fluency comes from doing it, not from reading about it.
Frequently asked questions
What is a Claude Code cheat sheet?
A Claude Code cheat sheet is a quick-reference of the slash commands, core concepts, and workflow you use most, written so you can screenshot it and stop looking things up. This one describes each command's purpose in plain language rather than listing flags that change between versions.
What are the most useful Claude Code slash commands?
The ones you reach for daily are /help, /clear, /compact, /init, /review, /agents, /mcp, /config, /cost, /model, /resume, and /skills. Run /help first whenever you are unsure, since it always reflects the exact version you have installed.
What is the difference between a skill, a subagent, an MCP server, and a hook?
A skill is reusable know-how the agent loads on demand. A subagent is a focused helper for a contained job. An MCP server connects Claude Code to an outside tool or data source. A hook fires an action automatically at a set point. Use a skill to repeat a task the same way, a subagent for a separable piece, an MCP server to reach outside data, and a hook for behaviour that should happen every time on its own.
Do I need the terminal to use Claude Code?
No. The desktop app runs the same agent without any terminal, which is why the CCC build flow starts there. You open the app, point it at a folder, give it a goal with a finish line, tell it how to verify, and review the result.
How do I get better results from Claude Code prompts?
Use three habits: put the goal in the first sentence (BLUF), name the files or folder in play to scope the work, and give the agent a way to check itself such as a test or expected output. Most disappointing runs come from a goal with no finish line and no check.
Why does this cheat sheet not list command flags?
Flags and exact syntax shift between Claude Code versions, so printing them would make the cheat sheet wrong over time. Describing each command's purpose stays accurate, and /help always shows the precise options for the version you have installed.
Last reviewed by David Iya on June 25, 2026
