title: "AI Skills for Gemini CLI & Codex CLI" description: "Configure coding standards and AI behavior in Gemini CLI and Codex CLI." slug: "gemini-codex-cli-skills" category: "skills" updatedAt: "2026-03-05T00:00:00.000Z" faqs:

q: "What is GEMINI.md?" a: "GEMINI.md is a project-level instruction file placed in your .gemini/ directory. Gemini CLI reads it to understand your project conventions."
q: "Does Codex CLI support custom instructions?" a: "Yes. Codex CLI reads instructions from .codex/instructions.md and respects custom configuration files."

AI Skills & Rules

2026

Updated Mar 5, 20265 min read

gemini cli

codex cli

skills

instructions

AI Skills for Gemini CLI & Codex CLI

Overview

Both Google's Gemini CLI and OpenAI's Codex CLI support custom coding standards through instruction files. Here's how to configure them.

Gemini CLI — GEMINI.md

Gemini CLI reads a GEMINI.md file from your .gemini/ directory:

# Project: E-Commerce API

## Tech Stack
- Runtime: Node.js 20 with TypeScript
- Framework: Express.js
- Database: PostgreSQL with Prisma ORM
- Auth: JWT with refresh tokens

## Coding Standards
- Use async/await over callbacks
- All functions must have JSDoc comments
- Use Zod for request validation
- Follow RESTful conventions for API endpoints

## Testing
- Jest for unit tests
- Supertest for API integration tests
- Minimum 80% code coverage

File locations:

Project-level: .gemini/GEMINI.md (recommended)
Global: ~/.gemini/GEMINI.md

Codex CLI — Instructions

OpenAI's Codex CLI uses custom instruction files:

# Codex Instructions

## Architecture
- Clean Architecture with repository pattern
- Dependency injection using tsyringe
- Domain-driven design for business logic

## Error Handling
- Custom AppError class for all application errors
- Never catch and swallow errors silently
- Log all errors with structured JSON logging

## Security
- Input validation on all endpoints
- Parameterized queries only (never string concat)
- CORS configured for specific origins

File location: .codex/instructions.md or ~/.codex/instructions.md

MCP Configuration

Both tools support MCP servers:

Gemini CLI

// .gemini/settings.json
{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["@modelcontextprotocol/server-github"],
      "env": { "GITHUB_TOKEN": "${GITHUB_TOKEN}" }
    }
  }
}

Codex CLI

// .codex/config.json
{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["@modelcontextprotocol/server-postgres"],
      "env": { "DATABASE_URL": "${DATABASE_URL}" }
    }
  }
}

Comparison

| Feature | Gemini CLI | Codex CLI | |---------|-----------|-----------| | Skills file | GEMINI.md | instructions.md | | Location | .gemini/ | .codex/ | | MCP config | settings.json | config.json | | Provider | Google | OpenAI |

Was this guide helpful?

Last updated: March 5, 2026