Skip to content
Back to search

threejs-game-skills

Agent skills for building playable, polished Three.js browser games with gameplay, AAA-style graphics, UI, QA, and optional AI-generated 3D, image, and audio assets.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/majidmanzarpour--threejs-game-skills/badge.svg)](/listings/majidmanzarpour--threejs-game-skills)

HTML

<a href="/listings/majidmanzarpour--threejs-game-skills"><img src="/api/listings/majidmanzarpour--threejs-game-skills/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Install all skills for Codex:

npx skills add majidmanzarpour/threejs-game-skills --skill '*' -a codex -g -y

Install all skills for Claude Code:

npx skills add majidmanzarpour/threejs-game-skills --skill '*' -a claude-code -g -y

If your installed `skills` CLI does not support one of these targets, install from a cloned checkout instead: `./install.sh --codex` for Codex, `./install.sh --claude` for Claude Code, or `./install.sh --all` for both.

For local development from a cloned checkout:

./install.sh --codex
./install.sh --claude
./install.sh --all

The local installer copies `skills/` into the selected agent skills directory. It skips same-named skills unless you pass `--force`, and it never removes unrelated user skills unless `--prune-managed` is explicitly requested.

./install.sh --codex --force

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 268 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Three.js Game Skills
  2. Self-contained Codex and Claude Code skills for building playable, polished Three.js browser games. Install the skills, then ask your agent to use `threejs-game-director`; the director routes gameplay, graphics, UI, asset generation, audio, debugging, and release verification without requiring users to choose every specialist skill manually.
  3. The package includes the runtime materials agents need: `SKILL.md` files, references, helper scripts, and a Vite + TypeScript + Three.js scaffold bundled inside the relevant skill folders. Generated games ship with deterministic test hooks, a seeded RNG, and Playwright templates for smoke tests, visual-regression baselines, and bot playtests so agents can verify their own work end to end.
  4. One shared pack of nine skills serves Codex and Claude Code. The director preserves your scope and art direction, uses specialist guidance automatically, delegates independent work when the runner supports it, and scales verification to the change. Full games get a complete production pass; a small HUD fix stays a small HUD fix. No particular model or paid API key is required.
  5. Created by [Majid Manzarpour](https://x.com/majidmanzarpour).
  6. ## Demos
  7. | Game | Video | Play |
  8. | --- | --- | --- |
  9. | Neon Ridge Drift | [Watch on X](https://x.com/majidmanzarpour/status/2064565389036540327) | [ridgedrift.netlify.app](https://ridgedrift.netlify.app) |
  10. | Championship Snooker Arena | [Watch on X](https://x.com/majidmanzarpour/status/2064673249129071096) | [snookerarena.netlify.app](https://snookerarena.netlify.app) |
  11. | Starship Dogfight | [Watch on X](https://x.com/majidmanzarpour/status/2065065340510281888) | [starshipdogfight.netlify.app](https://starshipdogfight.netlify.app) |
  12. | Tide Singer | [Watch on X](https://x.com/majidmanzarpour/status/2065570428723007555) | [tidesinger.netlify.app](https://tidesinger.netlify.app) |
  13. | Ripcore | [Watch on X](https://x.com/majidmanzarpour/status/2066687620709544070) | [ripcore.netlify.app](https://ripcore.netlify.app) |
  14. ## Install
  15. Install all skills for Codex:
  16. ```bash
  17. npx skills add majidmanzarpour/threejs-game-skills --skill '*' -a codex -g -y
  18. ```
  19. Install all skills for Claude Code:
  20. ```bash
  21. npx skills add majidmanzarpour/threejs-game-skills --skill '*' -a claude-code -g -y
  22. ```
  23. If your installed `skills` CLI does not support one of these targets, install from a cloned checkout instead: `./install.sh --codex` for Codex, `./install.sh --claude` for Claude Code, or `./install.sh --all` for both.
  24. For local development from a cloned checkout:
  25. ```bash
  26. ./install.sh --codex
  27. ./install.sh --claude
  28. ./install.sh --all
  29. ```
  30. The local installer copies `skills/` into the selected agent skills directory. It skips same-named skills unless you pass `--force`, and it never removes unrelated user skills unless `--prune-managed` is explicitly requested.
  31. ```bash
  32. ./install.sh --codex --force
  33. ```
  34. ## Use The Skills
  35. After installing, open Codex or Claude Code in an empty project folder, or in an existing Three.js game you want to improve. Then prompt the agent with the outcome you want and name the director skill:
  36. ```text
  37. Use threejs-game-director to build a premium futuristic tower defense game from scratch.
  38. Automatically use the relevant gameplay, graphics, UI, asset generation, audio, debug,
  39. and QA skills. Build a playable loop first, then iterate until it passes browser,
  40. mobile, visual, UI, performance, and release checks.
  41. ```
  42. Both runners share the same `SKILL.md` files and auto-discover the skills once installed; the prompt above works in either:
  43. - **Claude Code** reads skills from `~/.claude/skills` and routes from each `SKILL.md` description. Invoke the director with `/threejs-game-director`, or just name it in a prompt — it loads the sibling skills itself.
  44. - **Codex** discovers global skills in `~/.agents/skills` through the public install above. Each skill's `agents/openai.yaml` supplies its display name and a short `$skill-name` kickoff prompt. Use `$threejs-game-director` or name it naturally. The checkout's local `--codex` installer also supports the legacy `~/.codex/skills` location; avoid installing duplicate copies in both roots.
  45. The agent should:
  46. - Load `threejs-game-director` first for broad game work.
  47. - Load sibling skills for gameplay systems, AAA graphics, UI, debug/profile, QA/release, 3D generation, image generation, and audio generation when the request calls for them.
  48. - Use the bundled scaffold internally when starting from an empty folder.
  49. - Create or update the game code in your project.
  50. - Establish art direction, gameplay camera scale, and a representative playable scene before expanding levels or asset families.
  51. - For substantial builds, preserve constraints, completed work, pending asset jobs, and next actions in a compact project note so corrections do not restart completed generation.
  52. - Run checks appropriate to the change: full games include production builds, real input, screenshots, canvas pixels, supported viewports, and motion evidence when animated models matter.
  53. - Put detailed evidence and the premium visual scorecard in a project artifact; keep the final answer focused on what works and remaining risks.
  54. Users generally should not need to run the scaffold or QA helper scripts directly. Those scripts are packaged so the skills can use them as part of the workflow.
  55. Background tools, subagents, native async execution, user-steering delivery, and reasoning settings depend on the host. Installing these skills does not enable an API feature or change your model settings. When delegation is unavailable, the agent works directly.
  56. ## Optional API Keys
  57. The core Three.js skills work without paid API keys. When keys are genuinely missing or credits are exhausted, the director continues with procedural/local assets and reports the affected limitation. Temporary failures trigger bounded recovery of the existing job, not an immediate downgrade or duplicate paid submission. Explicitly procedural or no-external-service requests remain procedural. Add keys only when you want the agent to generate external models, images, or audio.
  58. Never commit API keys or put them in browser-side game code. These skills use provider APIs from local agent tooling, then save generated assets into your game project.
  59. | Provider | Skill | Environment variable | Use cases | Key setup |
  60. | --- | --- | --- | --- | --- |
  61. | Tripo API | `threejs-3d-generator` | `TRIPO_API_KEY` | Text/image/multiview to 3D, game-ready GLB/FBX hero models, vehicles, props, buildings, weapons, textures, rigging, animation, stylization, mesh conversion, post-processing. | [Tripo quick start](https://platform.tripo3d.ai/docs/quick-start) and [Tripo API overview](https://www.tripo3d.ai/api). |
  62. | Gemini image API | `threejs-image-generator` | `GEMINI_API_KEY` | Concept art, image-to-3D source images, texture references, decals, skies, backgrounds, icons, logos, GUI art, title/menu art. | [Gemini API key docs](https://ai.google.dev/gemini-api/docs/api-key) and [Google AI Studio keys](https://aistudio.google.com/app/apikey). |
  63. | ElevenLabs API | `threejs-audio-generator` | `ELEVENLABS_API_KEY` | SFX, music tracks and loops, ambience loops, UI sounds, announcer lines, dialogue TTS, voice conversion, audio cleanup, game audio manifests. | [ElevenLabs quickstart](https://elevenlabs.io/docs/eleven-api/quickstart) and [API authentication](https://elevenlabs.io/docs/api-reference/authentication). |
  64. Set keys in your shell profile, then restart your terminal.
  65. macOS/Linux with `zsh` or `bash`:
  66. ```bash
  67. export TRIPO_API_KEY="..."
  68. export GEMINI_API_KEY="..."
  69. export ELEVENLABS_API_KEY="..."
  70. ```
  71. For `zsh`, put those lines in `~/.zshrc` or `~/.zprofile`. For `bash`, put them in `~/.bashrc` or `~/.bash_profile`.
  72. Windows PowerShell, current terminal session only:
  73. ```powershell
  74. $env:TRIPO_API_KEY = "..."
  75. $env:GEMINI_API_KEY = "..."
  76. $env:ELEVENLABS_API_KEY = "..."
  77. ```
  78. Windows PowerShell, persistent for your user account:
  79. ```powershell
  80. [Environment]::SetEnvironmentVariable("TRIPO_API_KEY", "...", "User")
  81. [Environment]::SetEnvironmentVariable("GEMINI_API_KEY", "...", "User")
  82. [Environment]::SetEnvironmentVariable("ELEVENLABS_API_KEY", "...", "User")
  83. ```
  84. After setting persistent Windows variables, restart your terminal, Codex, or Claude Code so the agent process can see the new environment.
  85. The director skill includes a credential probe that sources common shell profiles before deciding a key is missing. Run it from wherever you installed the skills:
  86. ```bash
  87. # Claude Code
  88. bash ~/.claude/skills/threejs-game-director/scripts/probe_asset_credentials.sh
  89. # Codex
  90. bash ~/.agents/skills/threejs-game-director/scripts/probe_asset_credentials.sh
  91. ```
  92. It prints `TRIPO_API_KEY=SET|MISSING` (and the same for Gemini and ElevenLabs) without ever printing key values.
  93. Provider notes:
  94. - Tripo is optional but useful for high-value 3D surfaces that procedural code alone rarely makes premium: hero vehicles, bosses, weapons, buildings, creatures, props, and textured GLB/FBX assets.
  95. - Gemini image generation is optional but useful before Tripo image-to-3D and for high-quality texture, sky, icon, logo, decal, and GUI sources.
  96. - ElevenLabs is optional but useful for making games feel finished through interaction SFX, music, ambience, UI feedback, voice, and cleanup.
  97. - Google also supports `GOOGLE_API_KEY`, but these skills standardize on `GEMINI_API_KEY` for clarity.
  98. - Use provider-side key restrictions and quotas where available. ElevenLabs documents endpoint scopes, credit quotas, and secret-key handling; Google recommends environment variables and is migrating Gemini users toward auth keys.
  99. ## Best Entry Points
  100. - Use `threejs-game-director` for complete games, major upgrades, premium polish, release-ready work, or anything broad.
  101. - Use `threejs-gameplay-systems` for mechanics, architecture, input, camera, physics, scoring, objectives, and game feel.
  102. - Use `threejs-aaa-graphics-builder` when screenshots look basic or the game needs stronger models, materials, lighting, VFX, world detail, or render polish.
  103. - Use `threejs-game-ui-designer` for HUDs, menus, overlays, responsive layout, safe areas, icons, touch controls, and text fit.
  104. - Use `threejs-debug-profiler` for black screens, runtime errors, loading issues, resize/mobile bugs, performance, draw calls, triangles, textures, and memory.
  105. - Use `threejs-qa-release` for production builds, browser verification, screenshots, canvas pixels, mobile checks, release risk reports, and static-hosting readiness.
  106. - Use `threejs-3d-generator` for Tripo API text/image-to-3D models, texture, rigging, animation, conversion, and GLB/FBX game assets.
  107. - Use `threejs-image-generator` for Gemini-generated concepts, image-to-3D inputs, textures, decals, skies, backgrounds, icons, logos, GUI art, and title/menu art.
  108. - Use `threejs-audio-generator` for ElevenLabs SFX, music, ambience, UI sounds, voice/TTS, voice conversion, cleanup, and Three.js audio integration.
  109. For most user-facing game requests, start with `threejs-game-director` and let it pull in the specialists.
  110. ## What Good Usage Looks Like
  111. For a new game:
  112. ```text
  113. Use threejs-game-director to create a AAA-inspired hover racing game from scratch.
  114. Make it playable, add premium track visuals, vehicle feel, HUD, SFX hooks, desktop and
  115. mobile controls, and run the full verification pass before reporting done.
  116. ```
  117. For visual upgrades:
  118. ```text
  119. Use threejs-game-director to upgrade this Three.js game from prototype visuals to
  120. premium browser-game quality. Use the AAA graphics, UI, image, 3D, audio, debug,
  121. and QA skills as needed. Include the visual scorecard and evidence from active-play
  122. screenshots.
  123. ```
  124. For an existing broken game:
  125. ```text
  126. Use threejs-game-director to debug and finish this Three.js game. First get it running,
  127. then improve gameplay feel, UI, graphics, performance, and release verification until
  128. the remaining risks are explicit.
  129. ```
  130. For asset-heavy games:
  131. ```text
  132. Use threejs-game-director to build a premium space dogfight game. Use threejs-image-generator
  133. for concepts, skies, decals, icons, and GUI art; use threejs-3d-generator for hero ships
  134. and weapons when credentials are available; use threejs-audio-generator for SFX and
  135. ambience. If generation is blocked, report the credential probe output and fallback plan.
  136. ```
  137. ## Expected Evidence
  138. For a complete game, the skills should gather evidence before claiming success. Narrow edits use the affected subset, not an unrelated full release audit:
  139. - `npm run build`
  140. - local browser run
  141. - browser console and page error check
  142. - Playwright screenshot
  143. - canvas nonblank pixel check, plus the inspector's measured metrics (color entropy, edge density, luminance contrast, render budget rows)
  144. - supported-viewport checks (desktop and mobile by default)
  145. - interaction check for the main control path
  146. - game design brief, core loop contract, and level/encounter plan for broad game creation
  147. - performance snapshot when graphics, assets, shaders, or post-processing changed
  148. - technical art budget target-vs-actuals when premium graphics changed
  149. - UI text-fit, overlap, safe-area, and touch-target checks when UI changed
  150. - visual scorecard with measured evidence for premium, AAA, showcase, or less-basic claims
  151. - visual test harness decision (screenshot baselines added/extended/skipped) for release-ready visual QA
  152. - bot playtest metrics for release-ready gameplay claims
  153. - generated asset paths and task IDs, or the credential probe output, when external assets are in scope
  154. - short motion captures for animated models: rig deformation, clip transitions, foot sliding, and attack/contact timing
  155. - a current-run manifest declaring required viewport/state captures and artifact paths, with acknowledged states and a matching run ID
  156. Premium/AAA claims should not rely on a static scene, placeholder cubes, generic stat-card HUDs, or unverified screenshots. The game should have an active playable loop and a filled visual scorecard.
  157. ## Skill System
  158. - `threejs-game-director`: main entrypoint for complete game builds: scope and quality bar, specialist routing, bounded delegation, continuity notes, asset recovery, current-run evidence.
  159. - `threejs-gameplay-systems`: playable loop, architecture, game design and level design, mechanics, entities, controls, camera, physics selection, and game feel (hitstop, screenshake, easing, impact feedback).
  160. - `threejs-aaa-graphics-builder`: visual scorecard with calibration anchors, technical art budgets, shader/material cookbook, asset architecture, models, materials, VFX, render polish.
  161. - `threejs-game-ui-designer`: HUDs, menus, overlays, responsive UI, icons, safe areas, UI states.
  162. - `threejs-debug-profiler`: scene/runtime/render bugs, mobile bugs, performance profiling, renderer metrics.
  163. - `threejs-qa-release`: browser QA, screenshots, canvas pixels with measured metrics, visual test harness, bot playtests, responsive checks, production build, release risk report.
  164. - `threejs-3d-generator`: Tripo API text/image-to-3D, texture, auto-rig, animation, conversion, download, checkpoint/resume, staged inspection, and Three.js import guidance.
  165. - `threejs-image-generator`: Gemini image generation for concepts, textures, decals, skies, icons, GUI art, and image-to-3D inputs.
  166. - `threejs-audio-generator`: ElevenLabs-backed SFX, music, ambience, UI sounds, voice/TTS, voice conversion, cleanup, and Three.js audio integration.
  167. ## Packaged Resources
  168. Installed skills are self-contained. They do not depend on root docs, root scaffolds, root prompts, or root checklists.
  169. - `skills/`: the full public package. Each skill owns its required `SKILL.md`, `references/`, `scripts/`, and `assets/`.
  170. - `skills/threejs-gameplay-systems/assets/threejs-vite-game/`: packaged game scaffold used by the skills when starting from an empty project. Ships deterministic test hooks (`__THREE_GAME_TEST_HOOKS__`), a seeded RNG, and `tests/` templates for smoke tests, visual-regression baselines, and bot playtests.
  171. - `skills/threejs-qa-release/scripts/inspect-threejs-canvas.mjs`: packaged browser/canvas inspection helper: pixel metrics and render-budget rows. `--manifest` captures every declared viewport/state in one browser and prints one summary line per capture; explicit `--state` captures require an awaited matching acknowledgment.
  172. - `skills/threejs-aaa-graphics-builder/assets/scorecard-anchors/`: calibration reference screenshots for the visual scorecard.
  173. - `skills/threejs-game-director/scripts/check_evidence.py`: `--manifest` validates a declared capture set and run ID without scanning historical reports. The legacy `--report` interface checks cited artifact files but cannot prove freshness or complete capture coverage. Neither replaces visual inspection or gameplay tests.
  174. - `skills/threejs-game-director/scripts/probe_asset_credentials.sh`: prints `KEY=SET|MISSING` for all three providers, sourcing the shell profile the agent process does not inherit.
  175. - `scripts/`: local validation helpers for maintainers.
  176. - `install.sh`: local installer for working on this checkout.
  177. ## Maintainer Checks
  178. Validate this workflow repository:
  179. ```bash
  180. npm install
  181. npm run check:scripts
  182. npm run validate:skills
  183. npm run test:helpers
  184. ```
  185. Maintainers can run packaged helpers directly when testing the skill package, but normal users should interact through agent prompts:
  186. ```bash
  187. python3 skills/threejs-gameplay-systems/scripts/create_threejs_game.py ../my-threejs-game
  188. node skills/threejs-qa-release/scripts/inspect-threejs-canvas.mjs --url http://127.0.0.1:5188 --mobile
  189. node skills/threejs-qa-release/scripts/inspect-threejs-canvas.mjs --manifest artifacts/evidence.json --url http://127.0.0.1:5188 --seed 12345
  190. python3 skills/threejs-game-director/scripts/check_evidence.py ../my-threejs-game --manifest artifacts/evidence.json
  191. ```
  192. The [manifest reference](skills/threejs-game-director/references/evidence-manifest.md) documents capture paths and the state-hook contract. The [workflow evaluation scenarios](skills/threejs-game-director/references/workflow-evaluations.md) compare observable behavior with fixed prompts, seeds, environments, and model settings. Helper tests use mocked provider responses without spending credits; passing them is not a claim of improved visual quality.
  193. The shared orchestration policy incorporates [OpenAI's model guidance](https://developers.openai.com/api/docs/guides/latest-model?model=gpt-6-astra) on instruction conflicts, delegation, momentum, and proportionate checks. [Native async tools](https://developers.openai.com/api/docs/guides/async-tool-calling) and [steering](https://developers.openai.com/api/docs/guides/steering) remain hosting-application capabilities, not requirements of this pack.
  194. ## License
  195. MIT. See [LICENSE](LICENSE).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.