Unity Agent Workflows

Use this skill as the AI operating system for Unity game work. It turns project-specific lessons into a reusable workflow: read the live repo first, derive the user's actual project structure, prove the real runtime owner, route code to the existing owner/layer, avoid hub growth, validate the smallest useful surface, and publish durable rules only to the artifact that owns them.

Start Here

1. Read the live project instructions first: AGENTS.md, README.md, architecture docs, or any repo-local agent file.
2. Check git status --short and preserve unrelated dirty files.
3. If the user says $unity-agent-workflows. Teach, run the Teach command: create/refresh UNITY_STRUCTURE.md as a short index and split focused maps by category. Do not make one huge all-project document.
4. Derive only the relevant live project structure before architecture claims. Read existing folders, namespaces, .asmdef files, scenes/prefabs, bootstraps, graphs, and docs only when they are relevant to the requested category. Do not impose this skill's sample structure on the user's repo.
5. Inspect relevant files before editing. Do not start from memory or nearest-name guessing.
6. If a graph exists, read it before architecture claims only when dependency/routing proof needs graph data: graphify-out/GRAPH_REPORT.md, graphify-out/wiki/index.md, graph.json, or equivalent.
7. Classify the task before touching files:
   • Runtime/visible bug -> prove owner chain.
   • Visible target alignment, interactive/visual target focus, spotlight, modal dimming, duplicate names, or "do not guess" -> runtime visible target lock.
   • New or expanded C# responsibility -> project-derived routing.
   • UI layout/readability -> UI workflow.
   • Visual source asset -> visual asset gate.
   • Gameplay tuning/content -> data-first content workflow.