Use when interface for World Labs Marble/LWM APIs (3D scene generation, spatial reasoning, robotics integration). Benchmarks spatial tasks (object permanence, physics prediction, affordance detection). Integrates with robotics. Triggers on \"spatial-intelligence-world-models\", \"spatial intelligence world models\", \"models\".
Scanned 9/19/2026
Install to Claude Code
npx -y skills add majinmagros/magros.ai-skills --skill spatial-intelligence-world-models --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Spatial Intelligence World Models?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/majinmagros-spatial-intelligence-world-models)More formats (shields.io, HTML) on the badges page.
---
name: spatial-intelligence-world-models
description: "Use when interface for World Labs Marble/LWM APIs (3D scene generation, spatial reasoning, robotics integration). Benchmarks spatial tasks (object permanence, physics prediction, affordance detection). Integrates with robotics. Triggers on \"spatial-intelligence-world-models\", \"spatial intelligence world models\", \"models\"."
metadata:
origin: ECC
---
# Spatial Intelligence / World Models
World Labs Marble/LWM: image→3D scenes, spatial reasoning, robotics planning. Detalhes em `references/`.
## When to Activate
- Generating 3D scenes from single images (Marble API)
- Spatial reasoning: permanence, physics prediction, affordances
- Planning robot trajectories (grasp, door opening, manipulation)
- Benchmarking spatial capabilities against standard tasks
- Apps needing 3D world understanding (AR/VR, robotics, simulation)
- Skip: CAD modeling, pure 2D gen → `fal-ai-media`, Blender inspection
## Core Principles
1. **Validate geometry** — never treat 2D→3D as a black box
2. **Simulate before executing** — physics check prior to robot motion
3. **Multi-view for precision** — single view is not enough for robotics
4. **Mind sim-to-real** — verify on hardware before deploying
5. **Parameterize robots** — URDF/MJCF, never hardcoded configs
## Example
```bash
export WORLD_LABS_API_KEY="your_api_key"
worldlabs generate-scene --image input.jpg --output scene.glb
worldlabs reason --scene scene.glb --query affordances --region all
```
## References
- `references/marble-api.md` — capabilities, auth, request/response schemas
- `references/benchmarks.md` — task suite, targets, runner flow
- `references/robotics.md` — platforms, door/sandwich pipelines
- `references/pipeline.md` — end-to-end phases, errors, anti-patterns, scripts
## Checklist
- [ ] API key configured; scene quality validated
- [ ] Affordances verified per object category
- [ ] Force/torque limits + collision avoidance in plan
- [ ] Benchmark targets met before hardware trials
- [ ] Sim-to-real verified; emergency stop path tested
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!