three-live
llms.txt198 starsChanged 5 days agoApache-2.0View on GitHub
On-device JIT compiler and Web Components turning declarative HTML and AI-generated Three.js (<script type="text/three">) into WebGL2 & WebXR on the Web, Google Filament & Jetpack XR SceneCore on Android, and Apple RealityKit & Quick Look on iOS/visionOS.
In this repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- examples/llms-full.txtllms.txt
- examples/llms.txtllms.txt
- llms-full.txtllms.txt
- three-live-authoringSkill
- xr-live-plugin-devSkill
The file
We split this llms.txt into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 7
Always include allow-client-scripts on any <three-live> or <xr-live> element that contains an inline <script type="text/three">.
- Very specific: line 9
Never instantiate THREE.WebGLRenderer, GLTFLoader, TextureLoader, ShaderMaterial, or requestAnimationFrame inside <script …
- Very specific: line 14
Verify, visually inspect, and share scenes headlessly via npx three-live-cli validate|render|convert|sim|link (with --qr for terminal QR …
≈ “Common” is an estimate: at least 1 in 5 of the 3,592 llms.txt files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
# Three-Live & XR-Live (`<three-live>` / `<xr-live>`)
> On-device JIT compiler and Web Components turning declarative HTML and AI-generated Three.js (`<script type="text/three">`) into WebGL2 & WebXR on the Web, Google Filament & Jetpack XR SceneCore on Android, and Apple RealityKit & Quick Look on iOS/visionOS.
## Golden Rules for LLMs & Coding Agents
- Always include `allow-client-scripts` on any `<three-live>` or `<xr-live>` element that contains an inline `<script type="text/three">`.
- Inside `<script type="text/three">`, `THREE` and `scene` (`THREE.Scene`) are pre-defined globals in an isolated Worker sandbox. Add objects via `scene.add(mesh)`.
- Never instantiate `THREE.WebGLRenderer`, `GLTFLoader`, `TextureLoader`, `ShaderMaterial`, or `requestAnimationFrame` inside `<script type="text/three">`. Load external `.glb` files via `<three-live src="...">` or `<xr-anchor src="...">`.
- Attach spatial traits via `mesh.userData.xr = { grab, scale, billboard, follow, anchor, physics, orbit, handles, dimensions, snapTo, snapSocket, gaze, explode, navigable, sync, particles, holdToAsk, worldUnderstanding, splatUrl, splat, bloom, outline, ik, text, audio, character, agent, integrity, zone, spawner, objective, audioCues }`.
- Procedural CAD, Instancing & IK helpers in `THREE`: `Shape`, `Path`, `ExtrudeGeometry`, `TextGeometry`, `CSG` (`union`, `subtract`, `intersect`), `InstancedMesh`, and `IKSolver` (`solveTwoBone`, `solveFABRIK`).
- Attach declarative `KHR_interactivity` behavior graphs via `mesh.userData.interactivity = { onSelect, onHoverIn, onHoverOut, onStart, onReceive }`.
- Scene quotas: max `1,000` meshes, `100,000` vertices, `300,000` indices, `16` textures (`2048x2048`), `3,000 ms` execution timeout.
- Verify, visually inspect, and share scenes headlessly via `npx three-live-cli validate|render|convert|sim|link` (with `--qr` for terminal QR codes) or connect the built-in Model Context Protocol (MCP) stdio server via `npx -y -p three-live three-live-cli mcp` (`three_live_validate`, `three_live_render`, `three_live_convert`, `three_live_sim`, `three_live_link`).
## Documentation & Guides
- [Complete Single-File LLM Reference (`llms-full.txt`)](https://google.github.io/three-live/llms-full.txt): Full concatenated authoring guide, `userData.xr` trait dictionary, spatial plugin reference, and copy-paste recipes.
- [LLM & Agent Authoring Guide (`docs/llm-guide.md`)](https://github.com/google/three-live/blob/main/docs/llm-guide.md): Copy-paste system prompt snippet, allowed `THREE.*` constructors, and permissive fallback shims.
- [`<three-live>` Reference (`docs/three-live.md`)](https://github.com/google/three-live/blob/main/docs/three-live.md): Attributes, `<model-viewer>` drop-in API, `KHR_interactivity` authoring, PBR Next, HDRI, animations, and Worker/iframe sandboxing.
- [`<xr-live>` Spatial Computing & Plugins Guide (`docs/xr-live.md`)](https://github.com/google/three-live/blob/main/docs/xr-live.md): Canonical `userData.xr` traits, all 14 spatial plugins (`head-follow`, `manipulation-handles`, `world-understanding`, `gemini-live`, `spatial-paint`, `depth-particles`, `splat-loader`, `spatial-measure`, `magnetic-snap`, `gaze-dwell`, `exploded-view`, `persistent-anchors`, `spatial-locomotion`, `shared-room`), the 5 `xr-live-sim` simulation plugins, `<div slot="spatial">` cards, depth/physics, 25-joint hands, and native mobile bridges.
- [Developer Cookbook (`docs/cookbook.md`)](https://github.com/google/three-live/blob/main/docs/cookbook.md): Copy-pasteable end-to-end recipes.
## Agent Skills & Repository Rules
- [Repository Agent Instructions (`AGENTS.md`)](https://github.com/google/three-live/blob/main/AGENTS.md): Architectural invariants, 3-target parity contract, and build/test commands.
- [Scene & App Authoring Skill (`skills/three-live-authoring/SKILL.md`)](https://github.com/google/three-live/blob/main/skills/three-live-authoring/SKILL.md): Progressive-disclosure skill for building 3D and XR apps.
- [Cross-Platform Plugin Dev Skill (`skills/xr-live-plugin-dev/SKILL.md`)](https://github.com/google/three-live/blob/main/skills/xr-live-plugin-dev/SKILL.md): Contributor workflow for adding traits and plugins across Web, Android, and iOS/visionOS.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.