In this article, I’ll share eight powerful maneuvers in Claude Code that almost no one talks about. These tips will completely change the way you use the tool and will undoubtedly double the quality of your output. Let’s dive in.
1. The Power of Bash Mode
Imagine you’ve just opened Claude Code, ready to start a new task. Before you begin, you need to install a library like Axios, Tailwind, or any other package. The typical approach is to ask Claude directly: “install axios for me.”
This seemingly simple request triggers a four-step process: Claude reads your prompt, interprets your intent to run a command, executes that command, and then explains what it did. This is a lengthy process that needlessly burns through your tokens for a straightforward task. Worse, in some scenarios, the AI might not recognize your project’s package manager. It might default to npm when you’re using pnpm, leading to errors and a frustrating back-and-forth that wastes even more time and tokens. We can’t complain about running out of tokens by the end of the month if we continue with this approach.
There’s a much better way.
Simply type an exclamation mark (!), and you’ll notice the interface switches to “Bash Mode.” Now, type the command you want to run.
!pnpm install axios
Claude will execute this command directly in the terminal without any interpretation or wasted tokens. The output appears right in your session, ready to be used as context in your ongoing conversation. It’s a simple, direct, and efficient method.
2. Stash Your Prompts, Save Your Sanity
You’ve just spent time crafting the perfect, detailed prompt for your next big task. Every detail is explained. But just as you’re about to send it, you realize you forgot to clear the session to start fresh, free from the context of your previous work.
The instinctive reaction is to select all the text, cut it, run the /clear command, and then paste your prompt back in. It works, but it’s clumsy.
Let me show you a much more elegant solution. Instead of the select-and-cut dance, just press Ctrl+S. Your prompt will instantly disappear, temporarily stashed away. You’ll see a “Stashed” confirmation. Now, you can run any other command you need—in our case, /clear.
Once you submit the /clear command, two things happen: the session is cleared, and your original, stashed prompt magically reappears in the input box. No more selecting, cutting, and pasting. It’s a small feature, but it’s a huge time-saver.
3. Master Background Tasks
You’ve finished your main task and now you need to run the test files in your project to ensure everything is working correctly and nothing has broken. Typically, you’d ask Claude to run the tests and then wait. If you have a lot of test files, this can take a significant amount of time, leaving you stuck.
The solution is to combine two of the tricks we’ve learned.
First, use Bash Mode to run the test command yourself, saving tokens. Second, instead of waiting, press Ctrl+B. This sends the task to the background, freeing up the main thread. You can immediately start working on other tasks, even sending them to the background as well.
Once the background task is complete, Claude will notify you with the results. You can keep your workflow moving without interruption.
4. Queue Up Your Commands
This time, Claude is working on a large task, and you know that as soon as it’s done, you’ll need to run one or more follow-up tasks. Instead of waiting around for it to finish or getting distracted and coming back later, you can create a command queue.
Simply send the first command and press Enter. Then, type the second command and press Enter again. You can continue this for as many tasks as you need.
By doing this, you’re telling Claude: “When you finish your current task, execute these next.” As soon as the initial job is done, it will automatically start the next one in the queue, and so on, until everything is complete. When you return, you’ll find all your tasks finished and waiting for you.
5. The “Rewind” Button for Your Project
Now, let’s get into something more serious. Imagine Claude is working on your project and makes a mess, deleting important files. Or perhaps the implementation it chose isn’t what you had in mind, and you preferred the previous version.
In this situation, Claude offers a fantastic feature. Simply press the Esc key twice. A menu titled “Rewind” will appear, showing a list of all the checkpoints from your session.
From here, you can select any checkpoint and choose what you want to restore: just the conversation, just the file changes, or both. Once you make your selection, your project will instantly revert to its state at that checkpoint. Deleted files will be restored, modified code will return to its original state, and newly added code will be removed. This is a critical feature you’ll find yourself needing often.
6. Never Lose a Session Again
You’ve been working on a complex task for hours. You’ve built up a massive, valuable context with Claude. Then, the day ends, and you need to shut down your computer. Or worse, the power goes out, or Windows decides it’s time for a surprise update and restarts your machine.
Don’t panic. Your session, your tokens, and all the context you’ve painstakingly built are not lost.
You can simply run the /resume command. This will bring up a list of all your previous sessions, from newest to oldest. You can select the session you want—for instance, the one we started this article with—and continue your work exactly where you left off.
Pro-Tip: To make finding sessions easier, get into the habit of naming them. At the beginning of each session, use the /rename command to give it a name related to the task you’re working on. This will make it incredibly easy to find and resume any session at any time.
7. The “YOLO” Mode: High Risk, High Reward
Here it is, the final trick and the one I use most often. I’ve saved it for last because it’s both powerful and potentially dangerous. This is the “YOLO” mode, which stands for “You Only Live Once.” That’s not a joke.
In normal modes like Planning or Auto-Accept, Claude will always ask for your permission before it touches any file—whether to add, delete, or modify it. But in YOLO mode, you’re telling it to go full-throttle. You give it complete trust to proceed without ever checking back with you.
Think of it like a nuclear option: incredibly dangerous if misused, but extremely effective when handled with caution. Here’s how to use it safely.
First, to activate the mode, you start a session with a special flag:
claude --dangerously-skip-permissions
You’ll see a confirmation that you are now bypassing permissions. To stay safe, follow these steps:
- Confirm Your Scope: Double-check that you have Claude open within the correct project directory.
- Switch to Planning Mode: Change the mode to “Planning.” Take your time here. Work with Claude to build a rock-solid plan that you are confident in.
- Review the Plan: Meticulously review every step of the plan. Once you are certain it’s perfect, select the “Bypass Permissions” option.
Now, let it run. Go do something else. When you return, the entire task will be complete. And if something goes wrong or you don’t like the implementation? You can simply use the double-Esc trick to rewind to the previous checkpoint.
Bonus Tip: If at any point you want to review the plan Claude is following, you can use the /plan command. To open the plan directly in your IDE, use /plan open.
These tricks should fundamentally improve your efficiency and effectiveness when working with Claude Code.