What are Claude Code commands? Claude commands are shortcuts that let you
quickly control the AI, manage sessions, and perform actions without typing a
long prompt. See https://code.claude.com/docs/en/commands [1]
Starting with /help
Start up Claude
|
> claude |
For commands type in /
then the command
then hit enter to execute the command.
Also its important to note that this tries to help you. You can use tab completion and you can also
use the up/down arrow keys to select your command when you see it.
/help
Type in /help
|
> /help |
That will describe how commands work
Press the right arrow
key once
This will show the tabs
and put you in the commands tab
press the up/down arrow key to go through each command, one at a time.
Here we can see the
/memory cmd
Note: You can’t go between tabs once you press down you have to go all the way back up to move between tabs.
The custom-commands are
built in skills from anthropic. Your custom skills will now show up here.
(There is more for me to learn here but for now let’s move on and go through a
few commands and try to understand them)
/btw
The /btw (by the way) command is used to ask a question while claude is currently thinking.
A side question has full insight into the current conversation, but the /btw questions never enters the conversational history. It also cannot read files or run any tools it only knows about the current conversation history.
So for example if I run
|
> /btw how many files are in this folder |
I get back this explaining that it cannot do it
Unless that was already in the past conversation.
For example if I first ran this prompt
|
> count how many files are in this
folder |
Then ran
|
> /btw how many files are in this folder |
It can now answer this question because the data is in context.
Now it could be wrong if files were added or removed since the last command… but its still useful.
Also /btw can ask general questions to claude that are not part of your session.
|
> /btw What is the capital of Colorado |
It can answer this.
It just can’t run any tools locally.
/btw is meant to be a lightweight way to ask side questions while claude is processing (/btw will not interrupt claude thinking). /btw will not add/pollute the conversation.
So.. if you are working on some code and want to know the capital of Colorado use /btw so the conversation does not get derailed (added to the context/conversation)
see
https://code.claude.com/docs/en/interactive-mode#side-questions-with-/btw [2]
/skills
The /skills command will list all skills
|
> /skills |
I only have 1 skill at the moment. Here you can see ~60 tok
This means it will take about 60 tokens.
Press t to sort by tokens (largest at top)
I can’t do this as I only have 1 skill at the moment 😊
You
can also see on the left hand side it says “on”
What does that mean?
It means its fully enabled! Means that
the skill can be invoked using /name-of-skill or it can be invoked if Claude
thinks its relevant based on your prompt.
You
can change this.
For example if you only want to invoke it via the command an not allow claude to use it when it may think it needs to you can add this to your SKILL.md file for the skill you want to change
|
> vi
~/.claude/skills/command-history-html/SKILL.md |
Add
|
disable-model-invocation:
true |
Now save that and run the /skill command again
It
will now show user-only. So to invoke it
you have to use the / command version
You can also do the opposite. Disable the / command and only allow it to be run by claude when it thinks it needs to be run
In the SKILL.md file add
|
user-invocable:
false |
Run /skills again
You can see the skill is on, but if you try to run the / command it will not work. I think this needs to be updated to show its not invocable.
/cost /usage
|
> /cost |
This will show you how much you have spent in this session.
This is not your total usage this month but just for this
session.
if you open a second terminal and run /costs in it you will see that they are
tracked differently (on their own)
Now /cost actually opens the same thing
as /status but on the Usage menu. You
can use the left, right arrow key to find some interesting info
/status
|
> /status |
This will bring up the stats menu
If you use the left/right arrow key you can check the other
tabs
Config, you can see all your configs and change them (it even has a nice little search)
Just move to an option and use the right/left arrow key to
switch it.
Once you change it press enter to save
Note!!
Changing a setting will be saved… but where?
You would think it would be saved in ~/.claude/settings.json (and some
are). But some are saved in
.claude/settings.local.json in your working directory. Why?
Well from what I am reading some are UI or experimental and are put in
this local file vs yours.
In practice… If you want it for yourself save it in ~/.claude/settings.json.
Now in practice there is an order of precedence… who wins if
both have the same setting.
.claude/settings.local.json beats
.claude/settings.json, which beats ~/.claude/settings.json
Then there is the Usage menu which is where the /cost
command brings you.
And finally the Stats menu
This gives you some fun info graphics.
It even has a submenu Press down key then right arrow key
/Color
Sets the color for this session
Some options are red, blue, green, yellow, purpose, orange, pink, cyan, and
default
|
> /color red |
Now the session is red
If you have multiple sessions open this could be a nice way
to track them visually.
/clear
This will start a new conversation with an empty context, and history.
|
> /clear |
The old context is not lost it is saved to disk and can
brought back using the /resume command
New session!
Why would you want to do this? Good
question…
A few ideas come to mind
1. You are done with the old session and want to start a new one
2. You want to ask a few unrelated questions and do not want
to pollute the current context. You could /clear ask your questions then
/resume (If its just one question or task I would use /btw)
So when you /clear a session where does its context/memory get saved?
It gets saved in ~/.claude/projects/<project folder>/
then inside it will have a .jsonl file per session you run.
This is a json file. The l stands for
line so one json object is there per line.
And…. I just learned what JSONL is see https://jsonltools.com/what-is-jsonl
[3]
OK that is enough for now.. There are many more commands I
need to learn in depth but this is a good start.
References
[1] Agent Skills overview
https://code.claude.com/docs/en/commands
Accessed 05/2026
[2] Side questions with /btw
https://code.claude.com/docs/en/interactive-mode#side-questions-with-/btw
Accessed 05/2026
[3] What is JSONL
https://jsonltools.com/what-is-jsonl
Accessed 05/2026
No comments:
Post a Comment