unity-explorer-mcp

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local development commands such as npm install, npm run start, and lsof to manage the scene preview environment and local dev server.
  • [EXTERNAL_DOWNLOADS]: Fetches 3D models and assets from reputable community sources like kenney.nl and itch.io. It also uses npx to install additional topic skills from the official decentraland/sdk-skills package.
  • [DYNAMIC_EXECUTION]: Utilizes a bundled shell script (scripts/screenshot.sh) containing an embedded Python script. This script is used locally to parse JSON-RPC responses and decode Base64 image data from the local Explorer server to save screenshots to disk.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests runtime logs and entity state from the Explorer environment. While this represents a surface for indirect instructions, it is a standard requirement for its primary purpose of debugging the developer's own scene code.
  • [PRIVILEGE_ESCALATION]: Noted use of chmod is limited to the skill's own scripts, and there are no instances of sudo or unauthorized system-level modifications.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 09:04 PM
Security Audit — agent-trust-hub — unity-explorer-mcp