scenario-roblox-expert

Pass

Audited by Gen Agent Trust Hub on Nov 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of the Python subprocess module to interact with system utilities and development tools. These include ioreg, vm_stat, and sysctl for monitoring GPU and RAM usage on macOS, pgrep for process management, and osascript for UI automation (closing Roblox Studio windows via Accessibility features). It also invokes external CLI tools like rojo, luau-lsp, selene, and stylua to manage project builds and perform code analysis.
  • [EXTERNAL_DOWNLOADS]: The script scripts/rbx_check.py downloads a Luau type definitions file (globalTypes.d.luau) from the official GitHub repository of the luau-lsp project. This download is used to provide accurate type-checking for Roblox-specific APIs and is performed using standard library functions.
  • [DYNAMIC_EXECUTION]: The skill facilitates the execution of Luau code within Roblox Studio. It uses a structured protocol in scripts/rbx_run.py to wrap user-provided Luau snippets with a prelude and epilogue that capture output and serialize results to JSON. This execution is performed through the built-in Studio MCP server's execute_luau tool or via the Roblox Studio CLI's RunScript task.
  • [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests and processes untrusted data from Roblox game files and engine states (e.g., instance trees, script sources, and console logs).
  • Ingestion points: Game tree data is read via search_game_tree and inspect_instance, and script content is read via script_read (referenced in procedures.md).
  • Boundary markers: The Luau execution protocol uses specific delimiters (<<RBXRUN>> and <<RBXEND>>) to isolate the execution results.
  • Capability inventory: The skill has capabilities to write files (rojo build), modify scripts in the engine (multi_edit), and execute arbitrary Luau code (execute_luau).
  • Sanitization: The skill relies on the Roblox engine's own sandbox for Luau execution and uses structured JSON for result transmission, though it does not explicitly sanitize the content of scripts before reading or editing them.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 5, 2026, 11:32 PM