Skip to main content

Commands

Commands are reusable slash commands that you can type in the DevoxxGenie input field to trigger predefined prompts. They provide a fast way to perform common tasks like explaining code, generating tests, or reviewing changes.

Renamed from "Custom Prompts" / "Skills"

This feature was previously called Custom Prompts, and briefly Skills. It is now called Commands.

The new Skills feature is a separate capability — Skills are LLM-activated SKILL.md files loaded from disk, not user-typed slash commands.

How Commands Work​

  1. Type a slash command in the input field (e.g., /explain)
  2. Optionally add arguments after the command
  3. Press Enter — the command's prompt template is sent to the LLM

Commands support a $ARGUMENT placeholder that gets replaced with whatever text you type after the command. For example:

/test write unit tests for the UserService class

If the command's prompt template is:

$ARGUMENT

Then $ARGUMENT gets replaced with write unit tests for the UserService class.

Commands are a pre-LLM text substitution macro — DevoxxGenie expands the template locally before sending the resulting text to the model. This is fundamentally different from Skills, which are activated by the LLM mid-conversation via a tool call.

Built-in Commands​

DevoxxGenie ships with several built-in commands:

CommandDescription
/testGenerate unit tests for the selected code
/explainExplain the selected code
/reviewReview the selected code and suggest improvements
/findSearch for code in your project using RAG
/helpShow available commands
/initGenerate a DEVOXXGENIE.md project description file

Using /find with RAG​

The /find command requires RAG to be enabled and activated. When you type /find authentication flow, DevoxxGenie performs a semantic search across your indexed codebase and returns the most relevant code snippets.

If RAG is not configured, you'll see a notification explaining how to set it up.

Using /help​

The /help command displays a summary of all available commands directly in the output panel without sending anything to the LLM.

Managing Commands​

You can add, edit, and remove commands from the settings.

Accessing Commands Settings​

  1. Open IntelliJ IDEA settings
  2. Navigate to Tools > DevoxxGenie > Commands

Adding a Command​

  1. Click the + (Add) button
  2. Enter a command name (without the / prefix)
  3. Enter the prompt template — use $ARGUMENT where you want user input inserted
  4. Click OK

Example: Create a /docstring command:

  • Command: docstring
  • Prompt: Write comprehensive JavaDoc for the following code: $ARGUMENT

Usage: /docstring with code selected in the editor.

Editing a Command​

Double-click an existing command in the table to edit its command name or prompt template.

Removing a Command​

Select a command in the table and click the - (Remove) button.

Restoring Defaults​

Click the Restore button to reset the command list back to the built-in defaults. This removes any custom commands you've added.

Commands vs. System Prompt​

Commands and the system prompt serve different purposes:

  • System prompt: Sets the overall context and behavior for the LLM across all conversations. Configured in Settings > Prompts.
  • Commands: Individual slash commands that trigger specific prompt templates for one-off tasks. Configured in Settings > Commands.

Commands vs. Skills​

CommandsSkills
Invoked byUser types /name argsLLM calls activate_skill(name)
TimingPre-LLM, before the request is sentMid-conversation, LLM-driven
Requires Agent modeNoYes
StorageIDE Settings (XML)SKILL.md files on disk
Best forQuick repeatable prompt templatesReusable agent capabilities and workflows

See the Skills page for the new SKILL.md-based feature.

Tips​

  • Keep command prompts focused on a single task for best results
  • Use $ARGUMENT to make commands flexible — the same command can handle different inputs
  • Create project-specific commands for patterns you use frequently (e.g., /service to generate a service class following your project's conventions)
  • Combine commands with file context — select code in the editor before running a command to include it in the prompt
  • Skills — LLM-activated capabilities loaded from SKILL.md files
  • Agent Mode — let the LLM use tools and skills autonomously
  • MCP Support — extend capabilities with external tool servers
  • Chat Interface — use commands directly in the chat panel