04:54Tristen O'Brien · Tutorial
Claude Just Dropped ULTRA CODE
A 5-minute explainer that breaks UltraCode into its two parts, demos 48 agents in parallel, then reveals the cost.
June 6thThe argument in one line.
Six native slash commands fix the five biggest Claude Code productivity leaks — wasted context, mid-task interruptions, invisible session costs, premature execution, and irreversible mistakes — and the same mechanism lets you extend Claude with commands you design yourself.
Read if. Skip if.
READ IF YOU ARE…
- You use Claude Code regularly and have felt it drift, blow your context, or go down the wrong path before you could stop it.
- You type long setup prompts at the start of every session and want a single keystroke shortcut instead.
- You want to monitor your context usage and API spend without leaving the terminal.
- You have tried /plan but did not know about /btw, /rewind, or /goal.
SKIP IF…
- You work with Claude.ai chat, not the Claude Code CLI — these slash commands are CLI-only.
- You already have a custom command workflow and are looking for deeper architectural content.
The full version, fast.
Claude Code ships with dozens of slash commands that most users ignore. The six that matter most address distinct failure modes: /clear resets a bloated context window; /btw lets you feed Claude a side-question without interrupting an active run; /statusline turns the bottom bar into a live cost and agent dashboard; /plan forces a research-and-review cycle before any code is written; /rewind (aliases: /undo, /checkpoint) rolls back both code and conversation to a named checkpoint; and /goal keeps Claude running autonomously until a second checker agent confirms the goal is met. On top of those, you can define a custom command in plain English and Claude will build the file for you.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:53
01 · What slash commands are
Concept intro: a slash command is a shortcut — one keystroke vs. a long typed prompt. Animated long-way vs. short-way sequence.
00:53 – 01:54
02 · /clear
Wipes the context window to a clean slate. Animated context-window diagram. Notes /resume can restore cleared sessions.
01:54 – 02:44
03 · /btw
Passes context or a question to Claude without stopping an active task. Hallway-conversation metaphor with animated robot mascot.
02:44 – 03:13
04 · /statusline
Configures the bottom status bar to show model, context percentage, API cost, and active agent count. One-time setup.
03:13 – 04:02
05 · /plan
Forces Claude to research the codebase and produce a plan for user review and approval before writing any code.
04:02 – 05:04
06 · /rewind (and /resume)
Rolls back code and conversation to a checkpoint. Aliases: /undo, /checkpoint. Distinguishes /rewind (current session) from /resume (cleared session).
05:04 – 05:50
07 · /goal
Keeps Claude working autonomously until a stated goal is verified complete by a second checker agent.
05:50 – 06:37
08 · Build your own command (/scope)
Custom command that front-loads five clarifying questions, enters plan mode, and waits for explicit approval before executing. Full prompt template in description.
06:37 – 06:51
09 · Bonus: /radio
Easter egg that opens Claude's YouTube radio station.
06:51 – 07:23
10 · Bonus: /powerup
Interactive lesson guide for discovering other Claude features. Subscribe CTA at close.
Lines worth screenshotting.
- /btw lets you feed Claude a question or context mid-task without stopping the active run — treated like a sidebar, not an interruption.
- The /rewind and /resume commands solve different problems: /rewind rolls back the current session, /resume restores a session you already cleared.
- /goal does not stop when it thinks it is done — it requires a second checker agent to verify completion before marking the task finished.
- Typing /clear between unrelated tasks saves real money: Claude stops burning tokens re-reading irrelevant prior context.
- Building a custom slash command requires no coding — you describe what you want to Claude in plain English and it writes the command file.
- The most powerful pattern from this video is front-loading ambiguity: a command that asks five clarifying questions before planning is cheaper than replanning after a wrong build.
- The /statusline command is a one-time configuration that surfaces live data — model, context percentage, dollar spend, active agent count — at a glance.
- /plan is not just a review step — Claude actually reads the relevant code before proposing the plan, so the plan is grounded in what already exists.
- The aliases /undo, /rewind, and /checkpoint all do the same thing — pick whichever word you remember and it works.
- A custom command that forces clarification before execution effectively multiplies the value of /plan by ensuring the plan starts from a well-scoped brief.
Six shortcuts that fix Claude Code's biggest friction points.
WHAT TO LEARN
Every slash command in this video targets a specific, named failure mode — and knowing which command matches which problem is the skill.
- Clearing the context window between unrelated tasks is not a convenience feature — it prevents Claude from wasting tokens re-reading irrelevant history and producing confused output.
- The /btw command solves a real workflow problem: you can pass Claude a question or new context without interrupting an active task, so long-running builds do not need to stop every time you have a thought.
- The /rewind and /resume commands are not interchangeable — /rewind rolls back the current session to a checkpoint, while /resume restores a session that was cleared entirely.
- /plan does more than pause before acting — it reads the relevant code first, which means the plan it produces is grounded in what already exists rather than what the model assumes.
- /goal hands off the completion decision to a second checker agent, which means the task is not marked done based on the primary agent's own assessment.
- The most durable lesson is the custom /scope pattern: any command that asks clarifying questions before planning, and plans before building, will systematically produce better output than one that starts immediately.
Terms worth knowing.
- Context window
- The working memory Claude uses during a session — all prior messages, code, and responses. When it fills up, Claude starts making errors or ignoring earlier content.
- /btw (slash btw)
- A Claude Code command that lets you pass a question or additional context to the model without pausing its current task — the answer appears as a sidebar, not a task interruption.
- /rewind
- A Claude Code command that presents a list of conversation checkpoints and rolls both the code and the chat back to whichever point you select.
- /resume
- A Claude Code command that lists previous sessions that were cleared with /clear, letting you pick one to continue from where you left off.
- /goal
- A Claude Code command that keeps Claude working autonomously until the stated completion condition is met, with a second checker agent verifying the result before the task is marked done.
- Checker agent
- A secondary agent that /goal spins up to independently verify that the primary agent actually completed the task before declaring it finished.
- Slash command
- A shortcut in Claude Code: type a forward slash followed by a keyword and Claude executes a predefined behavior without you needing to write out a full prompt.
Things they pointed at.
Lines you could clip.
02:21
“It's almost like pulling a coworker out into the hallway to get what you need and then stepping back into the meeting.”
05:44
“It also won't mark the task done until a second checker agent verifies that the goal is actually complete.”
06:03
“I literally just told Claude what I wanted, and it works something like this.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00Slash commands inside of Claude are incredibly powerful, and most people have no idea how much they're actually missing. In this video, I'm gonna break down exactly what they are, exactly what they can do, and then the six that I use every single day. And at the end, I'll show you a custom command that you can build yourself that made my cloud outputs 43% better.
00:20So let's jump right in. I do wanna start off by explaining why a slash command is so powerful, and here's the simplest way that I can put it. A slash command is a shortcut.
00:29Instead of having to write out a long request to Claude every single time, you can just type in forward slash and a command, and Claude knows exactly what you want it to do. You'll see what I mean by that here in just a second. And the good news is they're really easy to use in the terminal or the desktop app.
00:44You just have to type in slash. And the second that you do, you're gonna see a list of commands that are available for you to use. It's really just that easy.
00:51So let me give you a couple examples and jump into the first one that I wanna highlight, slash clear. So this one sounds so simple, but I love it so much.
00:59This is what it does. When you type in slash clear, Claude starts a brand new conversation with a completely empty context.
01:07And the reason why this is such a great tool is because . When you are finished with one task and then jump into something that's totally different, all of that old stuff is still sitting there in the context window. So if you're on a new task, Claude will waste credits rereading all of that context from your previous task and try to understand the new one.
01:29And, ultimately, Claude is just getting confused. So when you wanna start something new, you're just gonna type in slash clear, and you're gonna get a clean slate. And if you're like me and you find out later that you really wish that you wouldn't have done that and you wanna go back to the old conversation, well, the good news is you can.
01:45But we're gonna talk about that a little bit later. This alone is gonna save you a lot of credits and make your projects run much faster. And speaking of keeping your conversation going, that brings me to number two, slash b TW.
01:57Alright. So you probably guessed it already. BT w stands for by the way.
02:01And this is something I use all day long. Matter of fact, this probably is my most used command in my arsenal. And here's what it does.
02:08Let's say that you're deep into a task with Claude code, and you have a random question or you wanna provide more context to Claude. You normally have to stop Claude from what it's doing to give it the information, but not with slash b t w. When you type in slash b t w and give Claude more information or ask it a question, it is going to answer you on the side instead of stopping an important meeting to ask a question.
02:33It's almost like pulling a coworker out into the hallway to get what you need and then stepping back into the meeting. Now this may seem small, but let me tell you, when you start using it, you're going to use it all the time. Okay.
02:44So number three is slash status line. At the very bottom of your cloud code, there is a little status bar. With slash status line, you can tell it exactly what you want down there.
02:54For me, I like to have the model that I'm using. I like to know how much of the context window I've already used, how much real money I would be spending if I wasn't using my subscription. Then I also like to know how many agents are running at any particular time.
03:07But, really, you could put anything you want down there. Now this really is a set it up one time type of thing, but it's awesome. So give it a shot.
03:13Number four is slash plan. Now I've talked about this a dozen times, but if you want significantly better outputs, you need to use this.
03:21Here's what it does. By default, when you ask Claude to do a big task, it is going to do that big task right away. Matter of fact, it's probably gonna build something awesome.
03:30But as awesome as that might be, it may not be exactly what you were looking for. If Claude took the wrong approach, you wouldn't find out until it was completely done. When you use slash plan, Claude researches the code that you're using or the request task that you put in, and it's planning what to do before it actually starts.
03:48You get to read it. You get to tweak it, and then you get to approve it. And from there, it goes to work like it did before.
03:55Now this is powerful, but I do have something even more powerful that boosted my output almost 43, but more on that in a bit. Now number five has saved me so much pain that it's not even funny.
04:05It's called the slash rewind command, and it does what it says. When Claude goes down the wrong path or it's making a bunch of edits that you didn't like, you can use slash rewind.
04:15And when you type that in and you hit enter, you're gonna see a list of conversation points or checkpoints that you've had with Claude. And all you have to do is select the one that you wanna revert back to, and then it rolls back both your code and your chat to that exact point.
04:30You can also use slash undo or slash checkpoint if you wanna use either of those.
04:35They all do the same thing. And remember earlier when I told you that you could go back to a conversation that you cleared? Well, I do wanna talk about that really quick because a lot of people mix these two up.
04:43Slash rewind rolls you back inside of the task that you're currently working on right now. But if you used slash clear earlier and you wiped out the entire conversation and you wanted to go back to that previous conversation, you would use a different command.
04:57You would use slash resume. You type in slash resume, and then you get a list of past conversations that you can jump right back into. Before we go on to number six, uh, if you could, do me a quick favor.
05:07If this has been helpful so far, please hit subscribe, like the video, leave a comment. This is something that really helps out the channel, and I certainly appreciate it. But let's get back to number six, which is slash goal.
05:18So I literally use this on almost every project. With slash goal, you tell Claude your task, and then you tell it what done looks like. And then Claude will keep working until it hits that goal, and that's how simple it is.
05:30One of the best things about slash goal is that, yes, it does keep working until it hits the goal, but it also won't mark the task done until a second checker agent verifies that the goal is actually complete. And in my opinion, this is one of the most powerful commands that are out there. Okay.
05:46So those are just six of the couple dozen built in commands that Claude has. But the best part is you can actually build your own custom slash command if you want to. I mentioned earlier that there's one that improved my output by 43%.
05:59How do I know that it improved it by that much? It's because Claude told me that it did. And to make it, I literally just told Claude what I wanted, and it works something like this.
06:06When I call the slash command and I give that task, Claude is gonna ask me at least five clarifying questions to fully understand my task. It's going to plan things out for me. But those five clarifying questions gives it more context than it knew before.
06:20And because of that, it's able to give me this output that is significantly better. Matter of fact, the difference is night and day.
06:27I'm gonna put those exact instructions that I gave Claude down in the description so you can paste them, and you can have Claude build the same command for you if you want to or just use it as a template to build something else. And before we wrap up, I do wanna give you two quick bonus slash commands because I think they're a little fun.
06:42First, you can type in slash radio, and it's gonna take you to Claude's own radio station on YouTube. First of all, the music's kinda good, and it's just a fun command to try.
06:51The second one is slash power up. Now I love this one because it's there to help you learn. What it does is it gives you a really quick interactive lesson guide on other cool things that Claude can do.
07:02So it's just a fun way to keep learning, and you're using a slash command. And that's it. These are the six slash commands that I use every single day.
07:09But, again, there's so many more that you can choose from. So take some time to look through them and find the ones that work best for you, and better yet, just make your own. So if you like this video, hit subscribe, hit like, and take a look at my last video here, and I will see you in the next
The Hook
The bait, then the rug-pull.
Most Claude Code users have discovered the forward-slash menu by accident, scanned the list, and moved on. This video is the argument for why that was a mistake — and it opens by naming the six commands that do the most work.
Named ideas worth stealing.
05:50concept
Front-load ambiguity before execution
The /scope custom command forces five clarifying questions plus a plan approval step before any code is written — eliminating the most expensive mistake in AI-assisted coding: building the wrong thing.
Steal forAny task-delegation workflow where requirements are fuzzy at the start
00:00list
Six-command friction map
- /clear
- /btw
- /statusline
- /plan
- /rewind
- /goal
Each command is mapped to a specific failure mode: bloated context, mid-task interruptions, invisible costs, premature execution, irreversible mistakes, and incomplete autonomous runs.
Steal forTeaching or onboarding someone to Claude Code
How they asked for the click.
VERBAL ASK
04:56subscribe
“If this has been helpful so far, please hit subscribe, like the video, leave a comment.”
Mid-video CTA placed between /rewind and /goal — before the custom command reveal which is the most valuable content. Soft ask, low pressure.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
AFFILIATECommission earned if you click.
Visual structure at a glance.
