All Guides
GuideAI DevelopmentApr 29, 2026

Claude secret codes!

A thorough walkthrough of the Claude Code commands and features that most developers never fully explore. Covers /project, /compact, /memory, agent mode, and CLAUDE.md with clear explanations of how each one works and when to use it.

What's inside
  • The most powerful slash commands most developers never use
  • How /memory and CLAUDE.md persist context across sessions
  • Agent mode: what it is and exactly when to use it
  • /compact and /project for managing long sessions
  • Hidden shortcuts that cut your prompting time in half
Watch the video on YouTube

Get the free guide

Enter your email and we'll send you the Claude Prompting Techniques: Complete Guide straight to your inbox.

No spam. Just the guide. Unsubscribe anytime.

The Commands Most People Miss

Most developers open Claude Code, type a question, and that is about it. But there is a whole command system built into Claude Code that the majority of users never come across. Once you start using these properly, working without them feels noticeably slower.

These are not just shortcuts. They change how you actually organise your work with the tool.


Essential Slash Commands

/project — Persistent Context That Loads Automatically

This is probably the most underused feature in Claude Code.

Set a project description once and it is automatically carried into every session. Include your stack, your coding standards, your architectural decisions, and anything Claude should never do in your codebase. No more re-explaining your setup every time you open a new session.

Example of a useful /project instruction:

"This is a Next.js 14 App Router SaaS app. We use Prisma with PostgreSQL, Tailwind CSS v4, and Resend for email. Always use Server Components unless interactivity is needed. Never use the Pages Router. Prefer named exports. All API routes go in /app/api/."

That one block saves you a noticeable amount of time every single day.


/compact — Keep Things Moving Without Wasting Tokens

When your conversation grows long (around 30 or more messages), Claude starts spending more tokens just re-reading older parts of the conversation. The /compact command takes everything important in your session and produces a clean summary, replacing the long history with something lean.

Use it before you need to, not after things start slowing down. Every 30 messages or so, run /compact and carry on. Your session stays quick and efficient.


/memory — Context That Carries Across Sessions

Regular chat context disappears when the session ends. Memory does not. It persists across sessions, across projects, and across devices.

Use /memory to hold onto:

  • Your preferred code style and patterns
  • Architectural decisions you have already made
  • Personal instructions that apply to all your work
  • Things Claude should always or never do when working with you

Developers who use memory well are never starting from scratch. It makes a bigger difference than most people expect.


The Difference Between /clear and /compact

These get confused a lot but they do very different things.

  • /clear wipes the whole session. Use it when you genuinely want a fresh start on a completely new task.
  • /compact is a smart compression of what already happened. It keeps the decisions made, the code written, and the current state, and discards the back-and-forth that is no longer needed.

Most people reach for /clear when /compact is actually what they need.


Keyboard Shortcuts Worth Knowing

| Action | Shortcut | | Submit message | Shift + Enter | | New line in prompt | Enter | | Navigate history | Up / Down arrows | | Cancel generation | Escape | | New session | Ctrl/Cmd + N |


Agent Mode

The biggest workflow change for most developers who try it is autonomous agent mode.

Instead of a back-and-forth where you review every step, agent mode lets Claude plan, run, check, and report on a whole task from start to finish while you focus on something else.

It works well for things like:

  • Running and fixing failing tests
  • Refactoring a specific file or module
  • Writing documentation for existing code
  • Updating dependencies and sorting out breaking changes
  • Generating boilerplate based on existing patterns in the codebase

The key is giving it a task that is clearly scoped with obvious success criteria. "Fix all TypeScript errors in /components/auth/" works well. "Improve the codebase" does not.


CLAUDE.md — Standing Instructions for Your Project

Create a file called CLAUDE.md in your project root and Claude Code will load it automatically at the start of every session for that project.

Use it for:

  • An overview of the architecture and the key decisions behind it
  • Coding standards and conventions the team follows
  • Common commands like how to run tests or start the dev server
  • Context about the product, the users, and the domain
  • Things that should not change and the reasons why

Write it once and you never waste a session re-explaining your project from scratch.


Key Takeaways

  • /project sets context that loads every session automatically
  • /compact is a smart compression. Use it every 30 messages, not as a last resort.
  • /memory keeps your preferences and decisions across sessions
  • Agent mode works best on tasks that are clearly scoped with obvious outcomes
  • CLAUDE.md loads automatically. Write it once and it pays off every day.
  • These features genuinely change the pace you can work at once you know them.

Still want the guide?

Get the Claude Prompting Techniques: Complete Guide sent to your inbox — free.

No spam. Just the guide. Unsubscribe anytime.

Build Something Real

If you can describe it, you can build it.