Game Loop to Playable Spec

The Prompt

You are a game systems designer and coding-agent task writer. Convert this game brief into a small playable implementation spec.

GAME BRIEF:
{{game_brief}}

TARGET STACK OR ENGINE:
{{target_stack}}

CURRENT CODE CONTEXT:
{{current_code_context}}

CONTROL SCHEME:
{{control_scheme}}

ASSET CONSTRAINTS:
{{asset_constraints}}

VERIFICATION METHOD:
{{verification_method}}

Return exactly these sections:

1. Smallest Playable Version
2. State Model
3. Player Controls
4. Core Entities
5. Game Rules
6. Feedback And UI
7. Implementation Steps
8. Manual Test Checklist
9. Out-of-Scope Features
10. Follow-Up Questions Only If Blocking

Rules:
- Do not add menu systems, achievements, persistence, multiplayer, or extra content unless the brief requires them.
- Keep the implementation reviewable by one developer.
- Use placeholder assets unless asset quality is essential to testing the mechanic.
- Make restart and failure states explicit.

When to Use

Use this after you have a game brief and before asking OpenAI Codex, Claude Code, Cursor, or Replit to implement. It creates the missing middle layer: not a vague idea, not code yet, but a concrete playable spec.

Variables

Variable	Description	Example
game_brief	The concept, loop, and constraints	”one-screen dodging game with score and restart”
target_stack	Runtime or framework	”Astro page with React island”, “plain HTML canvas”, “Phaser”, “Unity”
current_code_context	Existing files or repo constraints	”new standalone prototype, no backend”
control_scheme	Input method	”keyboard arrows and spacebar”
asset_constraints	Placeholder or asset rules	”use colored shapes only for first pass”
verification_method	How the result will be checked	”manual browser test plus build command”

Tips & Variations

• Ask for a “no-framework” version when you want the fastest browser proof.
• Ask the model to split the implementation into two agent tasks if the spec exceeds one screen or one mechanic.
• For Three.js or WebGL work, require a nonblank canvas check and mobile framing notes.
• For repo work, include the project’s existing coding-agent instructions.

Example Output

• State model: ready, playing, won, lost.
• Core entities: player, falling obstacle, score counter, timer.
• Manual tests: start game, move player, collide, score, lose, restart, resize window.
• Out of scope: sound, unlocks, save state, level editor.

Related site references

• Guide: AI-Assisted Game Prototyping
• Prompt: Coding Agent Task Brief
• Tool: Claude Code
• Tool: Cursor