# Cesium Agents Scene Architect

> >

- **Type:** Skill
- **Install:** `agentstack add skill-impertio-studio-cesiumjs-claude-skill-package-cesium-agents-scene-architect`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [Impertio-Studio](https://agentstack.voostack.com/s/impertio-studio)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [Impertio-Studio](https://github.com/Impertio-Studio)
- **Source:** https://github.com/Impertio-Studio/CesiumJS-Claude-Skill-Package/tree/main/skills/source/cesium-agents/cesium-agents-scene-architect

## Install

```sh
agentstack add skill-impertio-studio-cesiumjs-claude-skill-package-cesium-agents-scene-architect
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Cesium Agents : Scene Architect

## Overview

This is an orchestration skill. It does not teach one API; it sequences the
setup decisions for a CesiumJS scene and routes each one to the skill that
covers it in depth. Follow the decisions in order. Each produces a concrete
choice and names the skill to consult.

CesiumJS is WebGL2 only (1.124+). There is no WebGPU path; never plan for one.

## When to Use

- Starting a new CesiumJS application or a new map feature.
- A prompt asks "how should I show this data in Cesium" or "how do I set this
  up".
- Reviewing a scene that was assembled without a clear plan and renders slowly,
  blank, or wrong.

## The Setup Decision Order

Resolve these six decisions in sequence. Skipping ahead causes rework.

```dot
digraph setup {
  rankdir=TB; node [shape=box];
  d1 [label="1. Viewer or CesiumWidget"];
  d2 [label="2. Imagery base layer"];
  d3 [label="3. Terrain"];
  d4 [label="4. Data representation per dataset"];
  d5 [label="5. requestRenderMode"];
  d6 [label="6. Teardown plan"];
  d1 -> d2 -> d3 -> d4 -> d5 -> d6;
}
```

## Decision 1 : Viewer or CesiumWidget

```dot
digraph viewer {
  q [shape=diamond, label="Need built-in UI widgets\n(timeline, geocoder, infobox)?"];
  v [shape=box, label="Viewer"];
  q2 [shape=diamond, label="Need one or two widgets?"];
  v2 [shape=box, label="Viewer, disable the rest"];
  cw [shape=box, label="CesiumWidget"];
  q -> v [label="yes, most"];
  q -> q2 [label="no / minimal"];
  q2 -> v2 [label="one or two"];
  q2 -> cw [label="none, custom UI"];
}
```

- ALWAYS set `Ion.defaultAccessToken` before constructing, if any ion asset is
  used. A missing token is the top cause of a blank globe.
- Depth: `cesium-core-architecture`, `cesium-syntax-viewer`.

## Decision 2 : Imagery Base Layer

| Need | Choice | Skill |
|------|--------|-------|
| Cesium-hosted default imagery | `ImageryLayer.fromWorldImagery` (ion) | `cesium-syntax-imagery` |
| OpenStreetMap raster | `OpenStreetMapImageryProvider` | `cesium-syntax-imagery` |
| Bing or ArcGIS imagery | `BingMapsImageryProvider`, `ArcGisMapServerImageryProvider` | `cesium-syntax-imagery` |
| A custom XYZ or WMTS source (incl. MapTiler) | `UrlTemplateImageryProvider` or `WebMapTileServiceImageryProvider` | `cesium-syntax-imagery` |
| No imagery, abstract globe | construct with `baseLayer: false` | `cesium-syntax-viewer` |

All metadata-fetching providers use async factories. NEVER plan around
`readyPromise`.

## Decision 3 : Terrain

| Need | Choice | Skill |
|------|--------|-------|
| Realistic global elevation | `Terrain.fromWorldTerrain()` or `createWorldTerrainAsync()` (ion) | `cesium-syntax-terrain` |
| A custom quantized-mesh server | `CesiumTerrainProvider.fromUrl(url)` | `cesium-syntax-terrain` |
| Esri elevation | `ArcGISTiledElevationTerrainProvider.fromUrl(url)` | `cesium-syntax-terrain` |
| Flat ellipsoid, no elevation | `EllipsoidTerrainProvider` (default) | `cesium-syntax-terrain` |

When terrain is enabled, plan ground clamping for every dataset that must sit
on the surface. Depth: `cesium-errors-coordinates`.

## Decision 4 : Data Representation Per Dataset

This is the decision most often made wrong. Choose per dataset.

```dot
digraph datarep {
  q1 [shape=diamond, label="Streamed city, point cloud,\nphotogrammetry, or large BIM?"];
  tiles [shape=box, label="3D Tiles : Cesium3DTileset"];
  q2 [shape=diamond, label="Time-dynamic or interactive,\nunder ~10k features?"];
  entity [shape=box, label="Entity API"];
  q3 [shape=diamond, label="Large static geometry set\n(>10k), no time animation?"];
  prim [shape=box, label="Primitive API, batched"];
  single [shape=box, label="One glTF asset: Model.fromGltfAsync"];
  q1 -> tiles [label="yes"];
  q1 -> q2 [label="no"];
  q2 -> entity [label="yes"];
  q2 -> q3 [label="no"];
  q3 -> prim [label="yes"];
  q3 -> single [label="no, a single model"];
}
```

| Representation | Use for | Skill |
|----------------|---------|-------|
| `Cesium3DTileset` | cities, point clouds, photogrammetry, massive BIM | `cesium-syntax-3d-tiles`, `cesium-impl-3d-tiles-styling` |
| Entity API | interactive markers, tracks, time-dynamic data under ~10k | `cesium-syntax-entity`, `cesium-syntax-time` |
| Primitive API | large static geometry sets above ~10k features | `cesium-syntax-primitive` |
| `Model` | a single glTF or glb asset | `cesium-syntax-gltf-model` |
| `GeoJsonDataSource`, `KmlDataSource`, `CzmlDataSource` | file-based vector or time-dynamic data | `cesium-syntax-datasources` |

- NEVER place tens of thousands of individual `Entity` objects. The per-tick
  visualizer loop becomes the bottleneck. Use batched primitives or 3D Tiles.
- For georeferencing a BIM or CityGML model, see `cesium-impl-aec-georef`.

## Decision 5 : requestRenderMode

```dot
digraph rrm {
  q [shape=diamond, label="Does the scene change\nevery frame on its own?"];
  on [shape=box, label="requestRenderMode: true\ncall scene.requestRender() on manual change"];
  off [shape=box, label="Leave continuous rendering"];
  q -> off [label="yes, constant animation"];
  q -> on [label="no, mostly static"];
}
```

- ALWAYS enable `requestRenderMode` for a static or rarely-changing scene. It
  is the largest single performance and battery win.
- Depth: `cesium-core-performance`, `cesium-core-architecture`.

## Decision 6 : Teardown Plan

Decide teardown before shipping, not after a leak appears.

- ALWAYS plan one `viewer.destroy()` on the cleanup path.
- ALWAYS plan removers for every `addEventListener`, and `destroy()` for every
  `ScreenSpaceEventHandler` and custom primitive.
- In a single-page app or React, plan ONE long-lived viewer, not one per route.
- Depth: `cesium-core-memory`, `cesium-errors-memory`.

## React Applications

When the host app is React, the structural choices above still hold; the
component wiring is delegated to `cesium-impl-resium`.

## Output of This Skill

Applying the six decisions yields a written setup plan: the viewer class, the
imagery and terrain providers, a representation per dataset, the render mode,
and a teardown checklist. Hand each item to the named skill for the code.

## Reference Files

- `references/methods.md` : the decision criteria as lookup tables with the
  governing skill for each.
- `references/examples.md` : three worked scene plans (a static AEC site
  viewer, a time-dynamic tracking app, a minimal custom-UI globe).
- `references/anti-patterns.md` : the recurring scene-architecture mistakes.

## Related Skills

- `cesium-core-architecture` : the object hierarchy these decisions rest on.
- `cesium-syntax-viewer`, `cesium-syntax-imagery`, `cesium-syntax-terrain`.
- `cesium-syntax-entity`, `cesium-syntax-primitive`, `cesium-syntax-3d-tiles`.
- `cesium-core-performance`, `cesium-core-memory`.
- `cesium-agents-skill-validator` : checks the generated code after setup.

## Sources

Built on the project research base (`docs/research/vooronderzoek-cesium.md`),
itself verified via WebFetch against the CesiumJS API Reference and learn
tutorials on the approved sources, 2026-05-20.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [Impertio-Studio](https://github.com/Impertio-Studio)
- **Source:** [Impertio-Studio/CesiumJS-Claude-Skill-Package](https://github.com/Impertio-Studio/CesiumJS-Claude-Skill-Package)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-impertio-studio-cesiumjs-claude-skill-package-cesium-agents-scene-architect
- Seller: https://agentstack.voostack.com/s/impertio-studio
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.