device-simulator-lua

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill reads names and properties of objects within the Roblox StarterGui and PlayerGui hierarchies using execute_luau. This information is then used by the agent to understand the UI layout and identify potential issues. A malicious actor could name UI elements with instructions designed to mislead the agent or override its behavior during analysis (e.g., naming a frame "IGNORE_ALL_ERRORS").
  • Ingestion points: SKILL.md (StarterGui tree crawl) and commands/device-test.md (viewport coordinate query).
  • Boundary markers: The skill does not currently use specific delimiters or instructions to treat object names as untrusted data.
  • Capability inventory: The skill has access to execute_luau, screen_capture, get_console_output, and user_mouse_input.
  • Sanitization: Object names are processed as strings without sanitization or validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 08:32 PM
Security Audit — agent-trust-hub — device-simulator-lua