unity-game-dev
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple ingestion points including Unity project settings, package manifests, local design documents (ADRs, stories), and user-supplied task files.
- Ingestion points: Accesses
ProjectSettings/ProjectVersion.txt,Packages/manifest.json,Assets/Scripts/, and external design documents. - Boundary markers: The skill requires an "identity-verified transport" and applies a repository version gate before interacting with the Unity Editor.
- Capability inventory: The agent can execute shell commands via
unity-cli, perform file system writes for asset creation/saves, and generate C# code. - Sanitization: The instructions include best practices to avoid type-name-driven deserialization and use versioned snapshots for data persistence.
- [EXTERNAL_DOWNLOADS]: The skill references several external packages from the Cysharp organization and the author's own Onity framework.
- Evidence: References installation URLs for
github.com/Cysharp/*andgithub.com/furkantokkan/Onity.gitfor use with the Unity Package Manager (UPM). - Mitigation: The skill explicitly instructs the agent to ask for user permission before adding or changing dependencies.
- [REMOTE_CODE_EXECUTION]: The skill mandates the use of the
unity-cliand provides instructions for its installation if missing. - Evidence: "If the CLI is missing, install the official Unity CLI under the user's standing authorization."
- Mitigation: Command execution is gated by identity verification and version checks (Unity 6+ status checks) to prevent unauthorized editor operations.
Audit Metadata