Robinwood

FREE TO DOWNLOAD AND USE

Skills

Three skills that teach an AI assistant how to build for OpenSimulator — terrain, placement and meshes. The same ones used on Robinwood.

WHAT A SKILL IS

Instructions your assistant
already knows to follow.

A skill is a folder of instructions and helper files. Install one, invoke it by name, and the assistant works to that method instead of improvising — recording the terrain contract before exporting, checking a rotated footprint before planting a tree, validating the actual exported file instead of the source scene.

Two of these were written alongside the guides. The third, $opensim-mesh-builder, is the working skill from the Robinwood project itself, copied over with its references.

STEP BY STEP

Installing them.

  1. Download

    Take the three skills individually below, or the complete package, which also carries the guides, images and the training terrain. Unzip it into a project folder your assistant can read.

  2. Ask your assistant to install them

    Open the folder as a local project and ask:

    Prompt to copy

    Review the Skills folder in this project. Install these skills into my user skills directory, preserving any newer local versions. Then confirm their names and how to invoke them.

    On the machine these were written on — Codex with GPT‑6 Astra — the skills live in ~/.codex/skills/. Other assistants use their own configured skill location; let yours put them where it expects. If a newly installed skill does not show up, start a new chat.

  3. Invoke one by name

    Name the skill explicitly at the start of the work. Each skill below has a starter prompt you can copy.

THE THREE SKILLS

What each one covers.

$opensim-terrain-llraw New with these guides
Download (6 KB)

Design or revise OpenSimulator terrain heightfields and export validated 13-channel LLRAW, float R32, previews and tile manifests.

Use it for: Islands, mainland, varregions, terrain maps, coastal grading or LLRAW conversion.

Starter prompt

Use $opensim-terrain-llraw to design a 256 meter island from my brief.

What is in the folder

  • SKILL.md
  • references/formats.md
  • scripts/terrain_io.py
  • agents/openai.yaml

Written up in full in Guide 1 — Terrain design and LLRAW export.

$opensim-terrain-object-placement New with these guides
Download (5 KB)

Add trees, grass, rocks, buildings and travel meshes to a terrain design using a reproducible placement manifest and a self-contained OAR.

Use it for: Landscape population, mesh libraries, terrain-aware placement, and preserving existing region content during additions.

Starter prompt

Use $opensim-terrain-object-placement to prepare a protected placement plan and object-only OAR.

What is in the folder

  • SKILL.md
  • references/placement-contract.md
  • agents/openai.yaml

Written up in full in Guide 2 — Adding objects to a terrain design.

$opensim-mesh-builder Robinwood project skill
Download (12 KB)

Build or revise Blender mesh assets with editable source, Collada upload files, UVs, material-face maps, LODs and simple collision geometry.

Use it for: Buildings, props, modular paths, signs, HUD meshes, or scriptable light assemblies intended for in-world import.

Starter prompt

Use $opensim-mesh-builder to build a complete mesh upload package from these references.

What is in the folder

  • SKILL.md
  • references/export-and-validation.md
  • references/robinwood-conventions.md
  • references/scriptable-surfaces.md
  • agents/openai.yaml

Written up in full in Guide 3 — Building and exporting complete meshes.

TRY IT ON SOMETHING SMALL

A terrain you can build
in a few minutes.

The exercise makes a 256 m island with a northwest hill and a southeast inlet, exports it, and decodes the result so you can check it. Run it from the extracted folder with a Python environment that has NumPy.

Download the exercise (313 KB)

python Examples/make_training_terrain.py
python Skills/opensim-terrain-llraw/scripts/terrain_io.py self-test
python Skills/opensim-terrain-llraw/scripts/terrain_io.py export \
    Examples/Training-Terrain/height.npy \
    Examples/Training-Terrain/Export --water 20
python Skills/opensim-terrain-llraw/scripts/terrain_io.py inspect \
    Examples/Training-Terrain/Export/terrain.llraw.raw \
    --width 256 --height 256

The bundled example measures a maximum LLRAW quantization error of about 0.0302 m, and its R32 matches the float32 master exactly. That validates serialization — not a live simulator load.

How these were checked

All 20 document pages were rendered and visually reviewed. All three skill manifests passed the skill validator. The terrain helper was exercised against valid, non-square and invalid inputs, and its example files were decoded and checked. No region was loaded or modified while the guides were prepared.

Example grade and avatar-clearance values in the guides are design and test inputs — not guarantees for every avatar, simulator or viewer.