# Unity Agent Workflows

Use for AI-assisted Unity work that needs live repo discovery, project-derived routing...

- Repository: https://github.com/hashgraph-online/awesome-codex-plugins (1479 stars)
- Category: Design
- License: MIT
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add hashgraph-online/awesome-codex-plugins --skill unity-agent-workflows -g -a claude-code -y`
Codex: `npx skills add hashgraph-online/awesome-codex-plugins --skill unity-agent-workflows -g -a codex -y`
Cursor: `npx skills add hashgraph-online/awesome-codex-plugins --skill unity-agent-workflows -g -a cursor -y`
Gemini CLI: `npx skills add hashgraph-online/awesome-codex-plugins --skill unity-agent-workflows -g -a gemini-cli -y`

## Description

Use for AI-assisted Unity work that needs live repo discovery, project-derived routing, runtime-owner proof, runtime-visible output hard stops, runtime numeric proof for repeated visible-output failures, state-step guards, multi-agent scope ownership, modular C#/asmdef safety, UI/scene/visual asset gates, data-first content changes, validation, cleanup proof, or durable workflow rules. Best when agents must prove the actual folder/module/scene/prefab/runtime owner before editing, especially runtime UI, generated assets, code graphs, tutorial/state flows, guided selection/action flows, overlay/dim source-bound mistakes, coordinate conversions, focus/highlight/marker/HUD alignment, predictive lead aiming, ballistic trajectories, spatial IoU verification, or repeated "fix still not visible" failures.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Unity Agent Workflows

AI contract for Unity work. Keep answers compact, but never remove exact paths, object/class names, commands, errors, validation, rollback notes, or proof requirements.

## Core Loop

0. No proof, no edit.
1. Treat this injected `SKILL.md` as authoritative; do not search for its own path unless missing or asked for diagnostics.
2. Read live project instructions narrowly: `AGENTS.md` if present, `git status --short`, and existing `UNITY_STRUCTURE.md` plus only the focused map that matches the task when available. Read `README.md`, architecture docs, `.agents/*`, or repo-local agent files only when the task cites them, needs architecture/new-file/asmdef/routing decisions, or maps are missing/stale.
3. Run `git status --short`; preserve unrelated dirty files.
4. If user says only `$unity-agent-workflows. Teach`, run Teach.
5. Derive only relevant live structure before architecture claims: folders, namespaces, `.asmdef`, scenes/prefabs, bootstraps, graphs, docs, `UNITY_STRUCTURE*`.
6. Classify task with `Task Gate`, then load required references before edits.
7. Inspect real owner files; never start from memory, nearest-name guessing, semantic similarity, or screenshot proximity alone.
8. Fill useful `Routing Card` fields for structure, visible output, state flow, multi-agent, or hub changes.
9. Patch the smallest proven owner/artifact.
10. Validate, sync mirrors if package skill files changed, then close out with proof.

## Task Gate

The rest: https://github.com/hashgraph-online/awesome-codex-plugins/blob/HEAD/plugins/AUN-PN/unity-agent-workflows/SKILL.md