agentleFS
sandraschisandraschi/worldlabs-mcp/skills/worldlabs/SKILL.md

worldlabs

Skill24 starsChanged 2 months agoMITView on GitHub

Generate navigable 3D Gaussian-splat worlds via the World Labs Marble API (text/image/video-to-world), browse the community gallery, and narrate scenes spatially through Spark 2.0

  • Reads credentials

3 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions3 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 3

    1. OpusGameLabs/game-creator★ 360same name
    2. JustinPerea/nebula-nodes★ 23same name
    3. worldlabsai/skills★ 9same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository5 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Agent (you): call the MCP tools directly.

    line 14

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

World Labs MCP — domain skill

MCP gateway to World Labs Marble 1.1 (world generation) + Spark 2.0 (streaming splat viewer with HRTF spatial audio). Backend bridge on :10865, webapp on :10864. 23 MCP tools, 2 Prefab cards, 2 resources, 2 prompts.

Agent vs human routing

  • Agent (you): call the MCP tools directly. Poll async work with get_operation; never claim a world exists before done=true with a world_id. Credit owners when reusing gallery prompts.
  • Human (Sandra): point at the webapp — Architect (/architect) for generation, Inbox (/inbox) for the action queue, Library (/library) for finished worlds, Spark viewer (/spark-viewer) for walkthroughs.

Tool catalog (real operation names)

Generate: generate_world_from_text, generate_world_from_image (is_panorama for 360 equirectangular), generate_world_from_multi_image (min 2 images + matching azimuths_deg), generate_world_from_video, upload_and_generate (local file end-to-end, 100 MB cap), plus the manual two-step prepare_media_upload → generate_world_from_media_asset. seed pins deterministic output; tags organises; disable_recaption stops the API rewriting prompts. Default model marble-1.1 (1500 credits, ~1-3 min); marble-1.1-plus auto-expands (1500 + 300/cube, slower).

Poll: get_operation (single poll, raises on failed ops), wait_for_world (blocking ≤90 s — suits marble-1.1 only; loop get_operation for plus).

Manage: list_worlds (paginated), get_world (SPZ/GLB/panorama/ thumbnail/CDN links are time-limited — download promptly), delete_world (irreversible, destructive).

Spatial (needs bridge :10865 + open Spark viewer): broadcast_spatial_notification (TTS at x/y/z via edge-tts), broadcast_spatial_audio (looping mp3/wav URL), place_world_tv (16:9 video plane), spawn_agent_avatar (default_agent = built-in GLB).

Meta: refine_with_local_llm (needs Ollama at OLLAMA_URL), generate_world_prompt (sampling-first expansion), worldlabs_help (detail quick/standard/verbose, topic generate/upload/poll/world/ spatial/meta), gallery_explore (see workflow 2), worldlabs_shutdown.

Resources: world://{world_id}, gallery://{tag}. Prompts: world-gen, spark-viewer.

Workflows

1. Text-to-world end-to-end

op = await generate_world_from_text(
    text_prompt="A misty Japanese garden at dawn, stone lanterns, koi pond",
    display_name="garden-dawn", model="marble-1.1", seed=42,
    tags=["garden", "morning"])
# op -> {operation_id, done: false}
final = await wait_for_world(op["operation_id"])
world = await get_world(final["world_id"])
# world -> splat_url (SPZ), mesh_url (GLB), panorama_url, thumbnail_url,
#           caption, world_marble_url (open in browser)

Rules: poll, don't assume; marble-1.1-plus needs a get_operation loop, not wait_for_world; download CDN assets immediately; record failures with their operation IDs (get_operation raises RuntimeError with the API reason).

2. Gallery prompt mining

hits = await gallery_explore(operation="search", query="cyberpunk", tag="sci-fi")
# -> compact {title, owner, seed, prompt, url} rows; tag "all" scans every tab
detail = await gallery_explore(operation="world", world_id="<uuid>")
# -> full SPZ/minimap asset URLs + marble.worldlabs.ai/world/{id} link

Reuse the creator's seed + prompt to reproduce a look, then vary. Credit the owner when sharing results. Ten showpiece-mined presets live in the webapp Style Gallery.

3. Spatial narration of a finished world

await broadcast_spatial_notification("Welcome to the garden", x=0, y=1.5, z=0)
await broadcast_spatial_audio("https://example.com/pond-loop.mp3", x=2, y=0, z=-3)
await place_world_tv("https://example.com/intro.mp4", x=-2, y=1.6, z=1)
await spawn_agent_avatar(x=1, y=0, z=1)  # default_agent GLB, surface-grounded

Preconditions: backend running (GET :10865/health → 200) AND the Spark viewer open (it consumes the narration SSE stream). Without the viewer the calls succeed but nothing is heard/seen — check that before debugging audio.

Troubleshooting

SymptomCause / fix
get_operation raises RuntimeErrorAPI-side failure; message carries the reason (credits, moderation, bad URL). Check balance at platform.worldlabs.ai/billing — web-app credits and API credits are separate pools.
wait_for_world times outNormal for marble-1.1-plus; switch to a get_operation loop.
Spatial tools "work" but silentSpark viewer not open, or bridge down. Verify :10865/health, then open /spark-viewer.
upload_and_generate rejects file100 MB cap; kind must be exactly image or video.
Asset URL 403sCDN links expire — re-fetch via get_world, download promptly.
refine_with_local_llm connection errorOllama not running; start it (default http://localhost:11434) or set OLLAMA_URL.

Configuration

VariableMeaning
WORLDLABS_API_KEYRequired. platform.worldlabs.ai/api-keys
WORLDLABS_BRIDGE_URLSpatial bridge (default http://localhost:10865)
WORLDLABS_LOCAL_PATHLocal asset folder (default ~/Downloads)
OLLAMA_URLLocal LLM (default http://localhost:11434)
PLEX_BASE_URL / PLEX_TOKENCinema Worlds integration
MCP_TRANSPORT=httpServe HTTP on :10865 instead of stdio

Docs: docs/ (SETUP, FEATURES, ARCHITECTURE, PROMPT_GUIDE, TTS, SPARK_V2, EXPORT_GUIDE, WEBXR, ONBOARDING), llms-full.txt (endpoint + env reference).

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.