Unity Skill

Control Unity Editor via OpenClaw Unity Plugin. Use for Unity game development tasks including scene management, GameObject/Component manipulation, debugging, input simulation, and Play mode control. Triggers on Unity-related requests like inspecting scenes, creating objects, taking screenshots, testing gameplay, or controlling the Editor.

This skill appears coherent for remote control of the Unity Editor, but it grants powerful capabilities inside your Editor/project (reading script files, executing arbitrary code via reflection, installing packages from git, deleting/moving assets, taking screenshots). Before installing: 1) Backup your project and test in a disposable project first. 2) Review extension/index.ts yourself (it will be copied to ~/.openclaw/extensions/unity). 3) Keep disableModelInvocation set to true unless you explicitly trust the remote gateway and want the agent to auto-call tools. 4) If you will use the Gateway (Telegram/Discord/OpenClaw channels), be aware results (console output, screenshots, script contents) can be relayed to remote channels — only enable remote access when you trust the channel/recipients. 5) If you do not want the skill to install code from external git URLs, avoid using the package.add with gitUrl option or disable the Bridge in environments where network-sourced packages are risky. 6) If you need higher assurance, review or run the extension in a restricted environment (no network access) and audit any added packages before accepting them.

Type: OpenClaw Skill Name: openclaw-unity-skill Version: 1.6.1 The skill is classified as suspicious due to the `unity_execute` tool's `script.execute` capability, which allows arbitrary C# code execution within the Unity Editor context (a significant RCE vulnerability). Additionally, the `package.add` tool enables installation of packages from arbitrary Git URLs, posing a supply chain risk. While these are powerful features for a development tool, they present high-risk attack surfaces if exploited, even though there is no direct evidence of malicious intent (e.g., data exfiltration, backdoor installation) in the provided files (`SKILL.md`, `references/tools.md`, `extension/index.ts`). The default `disableModelInvocation: true` setting in `SKILL.md` mitigates automatic exploitation but can be overridden.