Writing Custom AI Prompts and Commands in Remocode: A Practical Guide

# Writing Custom AI Prompts and Commands in Remocode

Remocode's built-in status and audit commands are powerful out of the box, but the real flexibility comes from custom commands. The Commands tab in AI Settings lets you define your own commands with tailored prompts, enabling workflows specific to your project and coding style.

Accessing the Commands Tab

Press Cmd+Shift+A to open the AI Panel, then navigate to AI Settings and click the Commands tab. Here you will find configuration for the built-in status and audit prompts, plus the ability to create entirely new commands.

Customizing Built-In Prompts

The status command generates a progress report based on recent terminal activity. Its default prompt works well for general use, but you can customize it to focus on what matters most to you. For example:

• ●Emphasize test coverage by adding instructions to report on test execution results
• ●Focus on API endpoints by asking the status report to list new or modified routes
• ●Track file changes by prompting for a summary of which files were created, modified, or deleted

The audit command's prompt determines what the security review looks for. You can extend it to include project-specific concerns like checking for hardcoded configuration values, verifying proper error handling patterns, or ensuring authentication middleware is applied correctly.

Creating Custom Commands

Custom commands follow a simple structure: a name and a prompt. When you invoke the command, Remocode sends the prompt along with relevant terminal context to your Chat Model and displays the result.

Here are practical examples of custom commands you might create:

Type Safety Check: Create a command called "types" with a prompt that asks the AI to analyze recent code changes for TypeScript type safety issues. Ask it to identify any uses of any, missing return types, and potential null reference errors.

Dependency Review: Create a "deps" command that reviews newly added npm packages or pip dependencies. The prompt should ask the AI to evaluate each new dependency for maintenance status, security history, and whether a lighter alternative exists.

Performance Review: A "perf" command with a prompt that evaluates code changes for performance implications. Ask it to flag potential N+1 queries, unnecessary re-renders, unbounded loops, or missing pagination.

Documentation Check: A "docs" command that reviews whether new functions and API endpoints have adequate inline documentation and whether the changes require updates to external documentation.

Writing Effective Prompts

Be specific about output format. Tell the AI exactly how you want results presented. Bullet points, tables, and severity ratings make output easier to scan.

Provide context about your project. Mention the framework, language, and conventions your project follows. A prompt that says "we use NestJS with TypeORM" produces more relevant analysis than a generic "check the code."

Limit scope. A prompt that tries to check everything produces shallow results. Focused prompts that target a specific concern produce deeper, more actionable feedback.

Include examples. If you want the AI to look for specific patterns, include an example in the prompt. Showing what a "bad" pattern looks like helps the model identify similar issues.

Combining Custom Commands with Agent Workflows

Custom commands pair naturally with AI coding agents. After Claude Code, Gemini CLI, or Codex completes a task, run your custom commands to validate the work from different angles:

• ●Run status for a progress summary
• ●Run your "types" command for type safety
• ●Run your "perf" command for performance review
• ●Run audit for security analysis

This creates a systematic review pipeline powered by AI, catching issues that might slip through a manual review.

Custom commands turn Remocode from a general AI coding tool into one that is specifically tuned to your project's needs and your team's standards.