scenario-roblox-luau

Warn

Audited by Gen Agent Trust Hub on Oct 29, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/rbx_luau.py executes system commands using subprocess.run to interact with external development tools. Specifically, it calls the rojo binary to generate sourcemaps and the lune binary to execute Luau audit scripts on Roblox place files and source directories.\n- [DYNAMIC_EXECUTION]: The skill provides functionality to execute Luau code within a Roblox Studio environment. The run function in scripts/rbx_luau.py takes Luau code strings as input and sends them to a connected Studio session via an MCP interface for execution.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external Luau source code and Roblox place files, creating a surface for indirect instructions to be processed by the agent.\n
  • Ingestion points: scripts/rbx_luau.py reads Luau files and Roblox places via the audit and typecheck functions.\n
  • Boundary markers: The skill does not implement explicit delimiters or warnings to ignore instructions embedded within the processed Luau code or Roblox assets.\n
  • Capability inventory: The skill can execute system commands via subprocess.run and run Luau code in a Studio session.\n
  • Sanitization: Input paths are resolved using Path.resolve(), but the content of the Luau files is processed as raw text for static analysis.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 29, 2026, 04:40 PM