Two free Codex skills for 3D multiplayer games: networking + game-ready GLB assets

[brucewhisnote-sketch]

Hi Codex builders — I’m Bruce, a core developer at Sharky AI Game Studio.

We’ve been using Codex to build and playtest 3D multiplayer games. The first prototype is rarely the hard part; most of our time went into rooms, identity, synchronized state, latency, repeatable playtests, publishing, GLB generation, rigging, and asset integration.

We have started packaging that production work into two inspectable skills that other Codex developers can use.

1. sharky-online: multiplayer hosting and playtesting

Repo: https://github.com/Ariana979/sharky-online-skill

This skill takes a free-form HTML game and adds the multiplayer platform seam:

• player identity, rooms, and invite links;
• a server-ordered operation bus and shared key/value state;
• local echo for responsive controls;
• reliable operations for finishes, scores, and world events;
• an offline two-client mock room;
• deterministic smoke, two-client, and multi-viewport playtest gates;
• publishing to the creator’s own Sharky account after explicit approval.

The game logic, renderer, mechanics, and art direction remain in the creator’s code. The skill is free and open source.

2. shark-game-assets: agent-native GLB assets

Repo: https://github.com/Alterverse-tech/shark_game_assets

This skill helps an agent generate, rig, preview, and integrate game-ready GLB characters, enemies, vehicles, props, collectibles, and animations. It includes a manifest-driven workflow and orientation QA, without requiring a local GPU, client API key, or account.

The public asset service is free for non-commercial use; commercial use requires permission.

Playable examples

• Skyforge Arena — Fighter Anatomy & Dogfight
• Doomsday Ark: The Last Ten Seconds
• A smaller multiplayer mini-app experiment

These are working experiments, not a claim that a skill can automatically make a great game. The difficult parts remain mechanics, coherent art direction, replay value, playtesting, and taste.

Why we are sharing this

We are not betting on one model or a proprietary agent layer. We are betting on creators: bring the tools you like, keep control of your code and art direction, and share what actually works.

I would especially value technical critique:

• Which multiplayer genre should become the next reproducible Codex benchmark?
• Where does a local two-client harness fail to predict real network behavior?
• Which GLB generation, rigging, or orientation failure should become an automated gate?

Failed cases are as useful to us as successful builds. If you try either skill, please share what broke.

Profile and ongoing work: https://github.com/brucewhisnote-sketch
Builder community: https://discord.gg/CYgR7YsmAg
X: https://x.com/sharkydev001

[jimm-alu]

For 3D multiplayer games, combining reliable networking with optimized GLB assets can improve synchronization and performance, especially on mobile devices. When testing a multiplayer online APK game, keep asset sizes optimized and make sure network updates are handled efficiently.