# godot-agent-vision

godot-agent-vision — a skill that gives agents eyes in Godot projects: cross-platform screenshot capture, WebP encoding, and visual review scored by a taste receptor system. Use when agents need to see UI, assets, or the editor.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode, Windsurf (Devin Desktop), Claude.ai · 287 installs · ★ 983 repo stars
- Author: thedivergentai
- License: LGPL-3.0 · policy=link-only · description only; the file stays with the author
- Source: https://github.com/thedivergentai/Game-Dev-Agentic-Skills/tree/HEAD/skills/godot-agent-vision/SKILL.md
- Updated: 2026-10-16
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/skills/godot-agent-vision

## What it does

godot-agent-vision is a host-side Python toolkit for capturing and analysing visuals in Godot projects. The main tool is a CLI `capture.py` with modes for assets, screen regions, editor windows, and viewports. Each screenshot is encoded to WebP with a budget (default short-edge 512px) before being passed to the model — no raw PNG walls in context.

The core is Taste Receptor Atlas v2: a table of 237 micro-receptors across 13 families (hierarchy, contrast, typography×41, affordance, space, color, icons, composition, light, motion, anti-slop, originality, identity), scored 0–2 each. The agent loads only relevant reference docs (e.g., `typography-sight.md` for text, `hier-contrast-receptors.md` for layout), not all at once, and returns a zero-list with fixes keyed to receptor IDs.

For editor work, a temporary bridge — a `@tool` EditorPlugin — is staged to `addons/_gdskills_agent_vision/`, captures the viewport to PNG, encodes to WebP, and removes itself after. Critical: the bridge must not persist in the project, and `.gdskills/` must be gitignored before first run.

## Requirements
- Works with: Godot

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add thedivergentai/Game-Dev-Agentic-Skills --skill godot-agent-vision -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/godot-agent-vision/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/godot-agent-vision/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add thedivergentai/Game-Dev-Agentic-Skills --skill godot-agent-vision -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/godot-agent-vision/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add thedivergentai/Game-Dev-Agentic-Skills --skill godot-agent-vision -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/godot-agent-vision/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install thedivergentai/Game-Dev-Agentic-Skills skills/godot-agent-vision
```

2. **Manually: put the files in place**
- File: `.github/skills/godot-agent-vision/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add thedivergentai/Game-Dev-Agentic-Skills --skill godot-agent-vision -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/thedivergentai/Game-Dev-Agentic-Skills --path skills/godot-agent-vision
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/godot-agent-vision/SKILL.md` — The legacy .windsurf/skills/ path still works

### OpenCode

**Skills CLI (Vercel)**
```bash
npx skills add thedivergentai/Game-Dev-Agentic-Skills --skill godot-agent-vision -a opencode
```

### Claude.ai

**Claude.ai**
- Zip the folder and upload it in Settings → Capabilities → Skills

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/thedivergentai/Game-Dev-Agentic-Skills/tree/HEAD/skills/godot-agent-vision/SKILL.md

## When to use it
- capture a screenshot of an asset or UI panel and deliver it to the agent as compressed WebP
- score interface quality by typography, contrast, and element placement
- inspect the 3D viewport of the editor without launching the game
- review layout and design without external tools
- automate visual QA at the agent level

## Example prompts
- Capture the Godot window and score its typography against micro-receptors
- Grab `res://ui/icons` as a contact sheet and rate color and icon semantics
- Inspect the editor viewport in 3D mode via the temporary bridge, then tear it down

## Similar
- [scenario-godot-vfx](https://affpapa.org/en/nevibecode/skills/scenario-godot-vfx.md) — skill · 469 installs
- [godot-shaders](https://affpapa.org/en/nevibecode/skills/godot-shaders.md) — skill · 12700 installs
- [godot-physics](https://affpapa.org/en/nevibecode/skills/godot-physics.md) — skill · 12400 installs
- [scenario-godot-pipeline-automation](https://affpapa.org/en/nevibecode/skills/scenario-godot-pipeline-automation.md) — skill · 444 installs
- [godot-mcp-enhanced](https://affpapa.org/en/nevibecode/mcp/godot-mcp-enhanced.md) — MCP server · ★ 106 repo stars
- [hera-godot](https://affpapa.org/en/nevibecode/plugins/hera-godot.md) — plugin · ★ 1596 repo stars

From the same set: [thedivergentai/game-dev-agentic-skills](https://affpapa.org/en/nevibecode/by/thedivergentai/game-dev-agentic-skills.md) — 100 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
