# Unity Ml Agents

> Use when working on Unity ML-Agents: training configs, BehaviorParameters, sensors, actuators, inference in builds, or com.unity.ml-agents packages.

- Skill: `rachitkumarrastogi/unity-ml-agents` (Agent Skill)
- Install (CLI): `npx skillmds@latest add rachitkumarrastogi/unity-ml-agents`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rachitkumarrastogi/unity-ml-agents/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: rachitkumarrastogi (https://skillmd.com/u/rachitkumarrastogi)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/rachitkumarrastogi/unity-ml-agents

---

# Unity ML-Agents

## Discovery (unity-mcp-server)

1. `get_ai_stack_summary` — packages, ONNX assets, script hits, training YAMLs
2. `get_ai_ml_package_inventory` — confirm `com.unity.ml-agents` version
3. `list_ml_agents_training_configs` — trainer YAML paths
4. `find_ai_related_scripts` — filter label **ML-Agents**

## Implementation checklist

- **Training**: YAML defines `behaviors`, `trainer_type`, hyperparameters; keep configs in repo root or `config/`.
- **Inference**: Prefabs/scenes need `BehaviorParameters` + trained `.onnx` in `Model` field; verify with `list_prefabs_with_component` for `BehaviorParameters` if serialized as component name.
- **Sensors/Observations**: Stack vector observations consistently between training and runtime.
- **Performance**: Prefer inference on worker thread; cap decision frequency for mobile.

## Common prompts

- *"Does this project use ML-Agents?"*
- *"List ML-Agents training configs and ONNX models"*
- *"Which scripts reference BehaviorParameters?"*