Slash Commands

Slash commands are reusable prompt templates that users can invoke from the AI assistant chat window by typing /<command-name>. Each command has a markdown body with an optional $ARGUMENTS placeholder that is replaced with the user’s input at invocation time.

Creating Commands

Commands are created from the Slash Commands page under More in the sidebar. Click Create Command to open the editor.

The editor uses a single markdown document with YAML frontmatter at the top:

    
    
  
---
name: "explain-code"
description: "Explain the provided code snippet"
argument-hint: "<code or filename>"
allowed-tools: "read_file"
---

Explain the following code in detail:

$ARGUMENTS

Frontmatter Fields

Field	Required	Description
`name`	Yes	Command name (lowercase letters, numbers, hyphens; max 64 chars)
`description`	Yes	Shown in the command picker dropdown
`argument-hint`	No	Hint text shown next to the command name in the picker (e.g. `<filename>`)
`allowed-tools`	No	Comma-separated tool names to auto-approve when this command runs

Body

The body is free-form markdown sent to the LLM as a user message. Use $ARGUMENTS where the user’s input should be inserted. If the body does not contain $ARGUMENTS, the command runs immediately on selection (no argument prompt).

Scope

Commands can be either:

User commands — owned by the creating user, only visible to them
Global commands — visible to all users with access (filtered by groups and zones)

Global commands can be restricted to specific groups. Both types can be zone-restricted.

Using Commands

In the AI assistant chat window, type / to open the command picker. Select a command or keep typing to filter. If the command accepts arguments, type them after the command name:

/explain-code def factorial(n): return 1 if n <= 1 else n * factorial(n-1)

Only active commands appear in the picker.

Managed Commands

Commands marked as managed are read-only — they cannot be edited or deleted from the UI. Managed commands are typically provisioned by an administrator or via the API.

Scripting

Slash commands can also be managed programmatically via the knot.slash_command library:

    
    
  
import knot.slash_command

# List all commands
commands = knot.command.list()

# Create a global command
knot.command.create(
    content='---\nname: "greet"\ndescription: "Greet by name"\n---\n\nSay hello to $ARGUMENTS',
    is_global=True,
)

# Update a command
knot.command.update("greet", content='---\nname: "greet"\ndescription: "Greet warmly"\n---\n\nWarmly greet $ARGUMENTS')

# Delete a command
knot.command.delete("greet")

See the knot.slash_command library reference for full API details.

Navigation