infrastructure-cursor-codex-integration

Cursor & Codex Integration

Purpose

Configuration files and prompt conventions for using Cursor IDE and OpenAI Codex with this game development skill ecosystem.

When to Use

Trigger: cursor, codex, .cursorrules, cursor rules, AI IDE, codex instructions, prompt conventions, context injection, AI coding setup

Prerequisites

• claude-code-game-workflow — understand the skill ecosystem first

Core Principles

1. Context is everything — AI coding tools perform better with clear project context
2. Rules prevent mistakes — .cursorrules catches common errors before they happen
3. Skill references in prompts — tell the AI which skill to follow for the current task
4. Genre-agnostic enforcement — rules should enforce the genre-agnostic constraint

Stack Rules for .cursorrules

The .cursorrules file sits at the repo root and is loaded automatically by Cursor. Use it to enforce the game-dev stack. Key rules to include:

# TypeScript strict mode — never `any`
# Runtime: Bun only — no npm/npx/yarn
# Backend: Elysia — never Express, Hapi, or Fastify
# Database: Drizzle ORM — no raw SQL, no Prisma, no TypeORM
# Genre-agnostic: shared engine code must never assume RPG/MMO/FPS
# Server-authoritative: all game logic on server, client is display only
# Narrative: quests/characters/lore require quest-narrative-coherence check FIRST

See templates/game-dev.cursorrules for the complete file with all ~25 rules.

Claude Code Integration

How to use Claude Code with this ecosystem:

1. Reference SKILL.md files with @ mentions — Claude Code can read skill files directly:

   @skills/game-dev/engineering/postgres-game-schema/SKILL.md implement the inventory schema
   

2. Create a project CLAUDE.md — use the template at claude-code-game-workflow/templates/game-project-claude.md. This file tells Claude Code which skills exist, when to use them, and project-specific constraints.
3. Set up hooks for automated skill routing — see claude-code-game-workflow/templates/claude-hooks-config.json for hooks that trigger the right skill based on file patterns (e.g., editing db/schema.ts triggers postgres-game-schema).
4. Import skill context inline:

   "@engineering/postgres-game-schema/SKILL.md implement the inventory schema"
   

   Claude Code will read the skill file and apply its patterns to the implementation.

Prompt Templates

Five concrete prompt templates for common game dev tasks:

1. New feature:

Read @skills/game-dev/engineering/[relevant-skill]/SKILL.md. Then implement [feature] following the patterns there. Stack: Elysia + Drizzle + Neon. Use Bun runtime. No raw SQL.

2. Narrative content (quests/characters/lore):

Read @skills/game-dev/narrative/quest-narrative-coherence/SKILL.md first. Then create [quest/character]. Follow the 5-step coherence check and output a coherence report before writing any content.

3. Schema change:

Read @skills/game-dev/engineering/postgres-game-schema/SKILL.md. Add [table] following the JSONB + indexes + soft-delete patterns. Include the Drizzle migration command.

4. Bug fix:

Read the game-state-sync boilerplate. The bug is [describe]. Apply a server-authoritative fix only — do not move any game logic to the client.

5. Code review:

Review this [file] against the patterns in @skills/game-dev/engineering/[relevant]/SKILL.md. Flag violations of: server-authoritative architecture, genre-agnostic constraints, TypeScript strict type safety, and Drizzle ORM usage.

Windsurf / GitHub Copilot

The same principles apply across AI IDEs:

• GitHub Copilot: .cursorrules content works in .github/copilot-instructions.md. Copy the rules file there and Copilot will apply them across the repo.
• Windsurf: supports .windsurfrules at the repo root with identical syntax to .cursorrules. Use the same templates/game-dev.cursorrules content.
• Rule priority: more specific rules override general ones. Put game-specific overrides at the top of the rules file.

Cross-References

• claude-code-game-workflow — ecosystem navigation
• All other skills — referenced in prompt patterns

Sources

• Cursor documentation
• OpenAI Codex best practices