A slash command is a shortcut for a prompt you send again and again. Instead of retyping “review my staged changes and flag anything risky” every time, you save that instruction once as a command - then type /review and 100xprompt expands it into the full prompt. Built-in commands cover the common jobs out of the box, and you author your own by dropping a single file into your project or home directory. The payoff: your best prompts become reusable, shareable, and discoverable - right where you work, with autocomplete to find them.

​
See how a slash command works

A slash command is a saved prompt template with a short name. Type / in the prompt and 100xprompt shows a menu of every available command with autocomplete. Pick one, add any arguments, and its template expands into a complete prompt that 100xprompt acts on.
The command itself is just a stored template. What makes it powerful is everything you put inside it: placeholders for arguments, references to files, and the live output of shell commands - all resolved at the moment you run it.

​
Use a built-in command

100xprompt ships with a set of ready-to-use commands. Type / to see them alongside any custom or plugin-provided commands.
CommandWhat it does
/initScans your project and creates or updates an AGENTS.md so 100xprompt understands your codebase, conventions, and workflows.
/reviewReviews your changes and flags issues. Accepts an optional target - commit, branch, or pr - and defaults to your uncommitted work.
/compactCondenses the current session’s history into a compact summary so a long conversation keeps going without losing the thread. Takes optional instructions on what to preserve.
/summarizeSummarizes the session so far, with optional guidance on focus.
/rememberReviews what 100xprompt has learned about your work and promotes durable project rules into your AGENTS.md.
/memoryReports the state of long-term memory. Pass consolidate to merge and prune stored memories.
/advisorGets a second opinion on the current approach from a stronger model.
Your installed Skills and Plugins can contribute their own slash commands too. A user-invocable Skill shows up in the / menu automatically, so /some-skill loads and runs it. See Skills and Plugins.

​
Build a custom command

A custom command is a single Markdown file. The file name becomes the command name - deploy.md becomes /deploy - optional frontmatter tunes how it runs, and the body is the prompt template.
deploy.md
---
description: "Draft release notes from recent commits"
agent: build
model: 100xprompt/pro
subtask: true
---

Write concise release notes for the following changes, grouped by
feature, fix, and chore.

Version target: $1
Recent history:

!`git log --oneline -20`

Reference the changelog format in @CHANGELOG.md.

​
Frontmatter fields

Everything in the frontmatter is optional. Include only what you need.
FieldPurpose
descriptionA short summary shown next to the command in the / menu. Write it so you can pick the right command at a glance.
agentRun the command as a specific agent, giving it that agent’s persona and tool permissions.
modelPin the command to a particular model, regardless of your session default.
subtaskWhen true, the command runs as a self-contained subtask so its work stays isolated from the main conversation.

​
Write the prompt template body

Everything below the frontmatter is the prompt that gets sent. Write it exactly as you’d phrase the request yourself - clear, specific, and outcome-first. This is the text that expands when you invoke the command.

​
Leave slots with argument placeholders

Make a command reusable by leaving slots for input you fill in at call time.

$ARGUMENTS

Captures everything you type after the command name as a single block. Best when the input is free-form.

$1, $2, $3…

Positional placeholders for structured input. $1 is the first word after the command, $2 the second, and so on.
---
description: "Open a focused bug investigation"
---

Investigate a bug in the $1 area with priority $2.

Details from the reporter:
$ARGUMENTS
Invoking /triage checkout high customers can't apply coupons fills $1 with checkout, $2 with high, and $ARGUMENTS with the full trailing text.

​
Pull in files and shell output

Two references let a command pull in real context at the moment it runs, so the prompt is never stale.
Prefix a path with @ to attach that file or directory to the prompt. 100xprompt reads it in when the command runs.
Review the API surface in @src/api and check it against @docs/api.md.
Paths are resolved relative to your project. Use @~/notes.md to reference a file in your home directory.
Shell references run real commands on your machine when the slash command is invoked. Only put commands you trust in a template, and keep them read-only where possible.

​
Choose where command files live

100xprompt looks for command files in two places. The format is identical in both - only the reach differs.

Project commands

Put files under .100xprompt/command/ in your project. These travel with the repo, so everyone on the team gets the same commands. Ideal for project-specific workflows.

Global commands

Put files under .100xprompt/command/ in your home directory. These are available in every project - perfect for your personal go-to prompts.
Tips:
  • Organize commands in subfolders to namespace them. A file at command/release/notes.md becomes /release/notes, keeping related commands grouped in the menu.
  • Keep team workflows in the project directory so everyone shares one source of truth.

​
Create your own command

Suppose you write the same standup update every morning. Turn it into a command so a single /standup drafts it from today’s commits.
1

Create the command file

In your project, add .100xprompt/command/standup.md. The file name sets the command name - this one becomes /standup.
2

Add frontmatter (optional)

Give it a description so it reads well in the menu, and pin a model or agent if you want:
---
description: "Draft a standup update from today's work"
---
3

Write the prompt template

Below the frontmatter, write the instruction - pulling in live context with a shell reference and leaving a slot for your notes:
Write a short standup update in three bullets:
yesterday, today, blockers.

Today's commits:

!`git log --oneline --since=midnight`

Extra notes: $ARGUMENTS
4

Invoke it

Back in 100xprompt, type / and start typing standup. Autocomplete surfaces your new command with its description. Select it, add any notes after the name, and press enter.
5

Watch it expand

The template resolves - the shell reference runs, your notes fill $ARGUMENTS - and 100xprompt receives the finished prompt and replies. Your command is now reusable and shareable via the repo.

​
Find commands in the menu

Every command - built-in, custom, Skill-provided, or plugin-provided - appears in the same / menu with autocomplete.
  • Type / to open the menu and see everything available.
  • Keep typing to filter by name; fuzzy matching finds partial matches fast.
  • Read the description shown beside each entry to pick the right one.
  • Add arguments after the name before you submit.
A well-written description is what makes a command discoverable months later. Treat it as the command’s headline.

​
Tips and limits

You’ll type these often. Prefer /triage over /investigate-a-reported-bug. Group with subfolders rather than long hyphenated names.
Start the body with what you want produced. A command is only as good as the prompt inside it - the clearer the instruction, the better the result.
If a command encodes how your team ships, reviews, or releases, keep it in the repo under .100xprompt/command/ so everyone shares one source of truth.
If a custom command shares a name with a built-in, the menu shows just one entry, so a collision can quietly replace what you expected. When the / menu doesn’t surface the command you had in mind, pick a distinct name - or namespace it in a subfolder - to keep both discoverable.
  • Skills: package reusable expertise that 100xprompt loads just in time and surfaces in the same / menu.
  • Terminal: where you type slash commands and watch them expand and run.
  • Workflows: chain commands and steps into repeatable, automated engineering workflows.

Skills

Package reusable expertise that 100xprompt loads just in time - and surfaces in the same / menu.

Terminal

Where you type your slash commands and watch them expand and run.

Workflows

Chain commands and steps into repeatable, automated engineering workflows.