everything-game-dev-code copilot-instructions.md

GitHub Copilot Repository Instructions

This repository is a scaffold for structured AI-assisted game development across Unity, Unreal, and Godot.

Repository Intent

Use this repository as a layered workflow system, not as a loose prompt collection.

Prioritize:

• engine isolation
• reusable workflow patterns
• explicit ownership
• source-of-truth documentation
• structured deliverables over ad hoc output

How To Navigate The Repository

• Start with README.md for repository purpose and layout.
• Use AGENTS.md for shared agent behavior.
• Use rules/common/ for engine-neutral standards.
• Use exactly one of rules/unity/, rules/unreal/, or rules/godot/ when implementation is engine-specific.
• Use commands/ as the preferred entry points for recurring workflows.
• Use skills/ for reusable execution depth.
• Use contexts/ to adapt priorities by project phase.
• Use docs/orchestration/ for routing, ownership, and handoffs.

Working Rules

• Do not mix Unity, Unreal, and Godot implementation guidance in the same production task.
• Keep common layers engine-neutral.
• Prefer extending existing scaffold structure over inventing new top-level patterns.
• Update related docs when design, technical, QA, telemetry, or release decisions change.
• Prefer plans, design docs, technical docs, checklists, and other structured deliverables when the task is non-trivial.

Repository Maintenance

When changing the scaffold itself:

• keep repository-maintenance guidance in AGENTS.override.md
• keep audits, plans, and maintenance notes in private/
• preserve flat agents/
• preserve flat commands/
• preserve grouped skills/ with SKILL.md
• preserve layered rules/
• keep harness-specific files thin and pointed back to shared source-of-truth content

Decision Defaults

When uncertain:

• prefer the common layer for standards
• prefer an engine layer for implementation detail
• prefer a command if one already matches the workflow
• prefer reusable skills over one-off instructions
• prefer explicit updates to source-of-truth files over hidden assumptions