Game-feel intelligence for AI-built games. Same mechanics, 10x the feel.
Ask an AI to build a game and you get something that works — and feels completely dead. No screenshake, no hit-stop, linear movement, snapping numbers, silent buttons. Every interaction technically correct, none of it satisfying.
This skill teaches Claude the craft of game feel: the trauma-based screenshake system, hit-stop timing tables, easing curves with exact parameters, squash-and-stretch ratios, particle budgets, and pitch-laddered audio — with playtested numbers, not vibes. Works for any engine or framework: canvas, Phaser, Unity, Godot, Pygame, anything.
git clone https://github.com/jayesh-bansal/game-juice.git ~/.claude/skills/game-juiceNext time you ask Claude to build a game — or ask "why does my game feel flat?" — it answers like a game-feel designer.
| File | Contents |
|---|---|
SKILL.md |
The 10 rules of juice + the 3-channel feedback principle + workflow |
data/recipes.md |
Exact numbers: shake trauma values, hit-stop durations, easing curves, particle counts, pitch ladders, camera recipes, and the juice-audit grid |
Every player action should answer in 3+ channels within 100ms: visual, audio, physical. One channel is noticeable. Three is satisfying. Five is juicy. This skill makes Claude ship the feedback with the mechanic instead of leaving it as "polish later" (which means never).
MIT