SKY CLUTCH is a fast, replayable aim / reaction mini game designed for a live event booth.
You have 60 seconds. Glowing sky orbs drift across the screen — click or tap them to pop them for points. Chain pops to build a combo multiplier up to x5. Dark storm clouds are decoys: click one and you lose 15 points and your combo. Rare Golden Clouds are worth big points and add 2 bonus seconds (max twice per game). When time runs out you get your score, accuracy, best combo, and a rank from Rookie up to SKY'S THE LIMIT — then one big PLAY AGAIN button resets everything cleanly for the next player.
Why it fits a booth:
- Under 3 minutes: a full visit (round + score screen) is under 90 seconds.
- Zero learning curve: one verb — click what glows. The whole game is explained in a single sentence on the menu; the first target appears within half a second of pressing PLAY.
- Clean reset: every run rebuilds all state from scratch, so players never inherit the previous player's score.
- Offline leaderboard: a daily top-10 board stored locally on the machine, plus a big-board display mode (?board=1) for a second monitor — no Wi-Fi, no backend, no account.
- No fail state: there are no lives, so a first-time player never fails in public in front of a queue.
Everything is original and made in code: the sky, the clouds, the mascot Nimbus (a small cloud pilot with goggles and a yellow scarf), and all sounds (synthesized with WebAudio). No third-party artwork, no logos, no copyrighted music. The game is a static site — vanilla HTML/CSS/JS, zero dependencies, zero network calls — so it runs from any static host or even straight from disk.
Controls: mouse, touch, or keyboard (Space / Enter to start and replay). Responsive from a 1280x720 booth monitor down to a portrait phone.
Built with IntelliJ IDEA Community Edition 2025.3 (opened, indexed, edited in IDEA; validated with Inspect Code; IDEA run configuration included in the repo). An AI coding agent assisted with writing the code and the automated Playwright verification; the AI workflow is documented in AI_SHOWCASE.md in the repository.