asset-transformer-toolkit
Fail
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: HIGHDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill is designed to generate C# scripts and execute them within a live Unity Editor session (via the
unity command script evalcapability as noted inSECURITY.md). The instructions inreferences/rulesets-and-actions.mdexplicitly guide the agent to 'Construct a script... to run as C# in a live Editor'. This grants the agent the power to perform runtime code generation and execution, which can be misused to perform unauthorized actions within the project scope. - [COMMAND_EXECUTION]: By executing generated C# code in the Unity Editor, the agent gains significant control over the local development environment. This includes the ability to create, modify, and delete project assets, change project settings, and interact with the Unity Editor's internal state, effectively bypassing standard user interface controls.
- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes external files and file paths.
- Ingestion points: External 3D file paths and types are ingested through functions like
CanImportFile(filePath)andCreateImporter(filePath, ...)inreferences/create-importer.md. - Boundary markers: The instructions do not specify any boundary markers or delimiters for these external inputs.
- Capability inventory: The skill possesses extensive capabilities including writing C# scripts, modifying
.assetfiles, and executing code within the Editor process. - Sanitization: There is no evidence of sanitization or validation of the content of the external files (e.g., metadata) or the file paths themselves, which could be exploited if an agent processes a maliciously crafted file.
Recommendations
- AI detected serious security threats
Audit Metadata