Effective communication is key to efficiency. By organizing your requests using a clear framework, you eliminate guesswork for Coplay and ensure the output—be it code, refactoring, or scene analysis—is always ready to use in your Unity project.
Every high-quality Coplay prompt should contain these three components:
1. Initial State (The Context)
Define where you are starting. Include the necessary background, the existing issue, or the component you plan to modify.
Example: “I have a third-person camera system using Cinemachine. The shake on impact is jarring.”
2. Desired Outcome (The Goal)
State exactly what you want the AI to deliver. Be specific about the functionality and the performance criteria.
Example: “Refactor the camera script to smoothly ramp up and then immediately decay the shake effect over 0.75 seconds to feel more natural.”
Tell Coplay where the change should happen and how the output should be structured.
Example: “Provide the full C# script for @CameraControl.cs with the new logic, ensuring all public fields are serialized.”
A great prompt is always focused: What is happening? ➡️ What should happen? ➡️ Where should it go?
Ensuring Project Focus
Coplay is project-aware, but you must direct its attention. Always reference specific files or objects to ensure the AI uses the correct context from your Unity project.
Always Reference Assets
Avoid generalized language. Use file pinning (like the @ symbol) or the in-editor selection tools to ensure Coplay is working on the correct scripts, prefabs, or scenes.
Effective: “Please update the PlayerAttack() function inside @MeleeCombatSystem.cs to…” Ineffective: “Fix the player’s attack script…”
Asking Coplay to perform tasks like “Fix all compilation errors” will yield poor results. Be precise: “Address the specific type error in the @HealthBarUI.cs script.”
Dealing with Subpar Outputs
If Coplay generates an incorrect or incomplete response, resist the urge to debug it conversationally within the same thread. Since you cannot edit past prompts, this only creates confusion.
Know When to Reset
If the AI misunderstands the core objective after the first attempt, it is always the most efficient choice to Start a New Thread.
- A clean start ensures the model is not polluted by the failed attempts and has a fresh contextual memory for your request.
- When starting the new thread, you may need to re-pin your assets and provide richer Initial State context to guide the AI correctly this time.
NEVER attempt to debug or fix a massive, incorrect AI output by conversationally pointing out dozens of errors. Start a New Thread instead.
