A nature-grounded world compiler for pixel art. Real physics, imaginary worlds.
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM lists this project for discovery and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
# Add to your Claude Code skills
git clone https://github.com/trinity-works/frostwindGuides for using ai agents skills like frostwind.
See how frostwind compares with popular alternatives.
Frostwind is a nature-grounded world compiler for pixel-art game worlds. Real physics, imaginary worlds. It turns explicit models of nature, life, culture, craft, and art direction into deterministic, inspectable scenes.
An LLM cannot see. Writing a 2D world, it guesses at every question a human artist answers by looking: how big is a deer next to a person, does a head read at this height, how far does a foot travel in one frame, what colour is snow in shadow at 62 degrees north. Each answer is plausible on its own, and together they produce a world where nothing agrees with anything.
Frostwind puts a derivation layer between intent and pixels. Physics settles what has a physical answer: 26 px is 1.70 m, a roe buck is 1.18 m, so he is 20 px. Craft settles what the medium constrains, like the row budget a head needs to read. Anything left over is declared as taste and applied consistently. Tests hold all three true as the world outgrows what anyone can keep in their head.
Nothing is drawn that can be derived. Nothing is chosen that can be measured. Nothing is claimed that cannot be checked.
Sprites store material, normal, occlusion, and thickness, never colour. Terrain is queried from analytic fields rather than painted maps. Reality and the pixel medium are both authorities; when they conflict, the legibility floor is explicit. An LLM may research and propose schema-checked tables, and pixels still pass through the same deterministic compiler: the model is the guardrail; the LLM writes tables, not pixels.
This is a physically grounded world model and reference implementation. It is not a general game engine or a runtime prompt-to-image system.
Play it in the browser, or build it from source with the three commands below. The code is open source under the MIT licence.
These are deterministic captures from the real renderer, not concept art. The capture commands are in tools/shot.mjs.
Use a current Node.js LTS release, npm, and a WebGL-capable browser.
git clone https://github.com/trinity-works/frostwind.git
cd frostwind
npm install
npx playwright install chromium # browser captures and frame tests only
npm run build
Open dist/frostwood.html directly; it is a self-contained page that works from
file://. The title screen offers Frostwood, Sun Meadow, Green Swamp, Amber Waste and
Amethyst Hollow;
Manifesto opens the project statement. Deep links skip the menu.
dist/frostwood.html?at=meadowdist/frostwood.html?at=meadow&poi=villageThe default scene is Frostwood. Its distant portal rebuilds the resident scene as Sun Meadow. Cabins, steadings, and swamp houses are outdoor props; doors do not cross into a separate room place.
The compiler is a five-stage pipeline:
Current proofs:
fields → affordance → cultural plan → candidates → pixels.The authored/procedural boundary is deliberate: nature generates continuous conditions and variation; authors state culture, landmarks, identity, and exceptions that no physical law can choose.
A world is data: plan rows and tables, in the order the add-world skill
(.claude/skills/add-world/SKILL.md) lists with the check for each step.
BIOMES row in src/biome.js:
scene, latitude, water model, traversal surfaces, ground remap, place ring,
settlement slots, cultures, and portals. src/worldgen.js
composes the stages in src/world/ from that row. Add a stage
branch only for a new model (a new kind of water, a new landform), keyed on
the plan.ROSTER in src/world/life-field.js.SCENES row in
src/scene.js with its biome, output ramps, haze treatment,
and water reflection window.npm run test:palette
to see which ramps moved, then look at a real frame before accepting it.Keep one sun, palette request, generator, atlas, and bake context per resident scene. If two worlds need different physical conditions, mak
frostwind is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by trinity-works. A nature-grounded world compiler for pixel art. Real physics, imaginary worlds. It has 80 GitHub stars.
frostwind's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/trinity-works/frostwind" and add it to your Claude Code skills directory (see the Installation section above).
frostwind is primarily written in JavaScript. It is open-source under trinity-works on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh frostwind against similar tools.
No comments yet. Be the first to share your thoughts!