unity-asset-management

Warn

Audited by Gen Agent Trust Hub on Apr 25, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: In references/runtime-checklist.md, the agent is instructed to use cargo run to execute the CLI tool if a global binary is not found on the PATH. This leads to the runtime compilation and execution of source code found within the local workspace.
  • [COMMAND_EXECUTION]: The skill relies on the unity-cli command-line tool to perform asset management tasks. It executes shell commands with JSON arguments to interact with the Unity editor.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion points: reads asset metadata and project files via manage_asset_database, Read, Grep, and Glob. Boundary markers: none observed in instructions. Capability inventory: shell execution via unity-cli, project file modifications, and network interaction. Sanitization: none documented for external asset data.
  • [DATA_EXFILTRATION]: The utility supports network connectivity to Unity instances via configurable host and port settings. This mechanism involves network operations that could be misconfigured to send project information to unintended destinations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 25, 2026, 02:58 PM