The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Roblox Studio that could contain instructions designed to influence agent behavior.
Ingestion points: Data enters the agent's context through script_read (reading existing game code), get_console_output (reading runtime logs), and search_game_tree (exploring game hierarchy), as documented in references/full.md.
Boundary markers: The instructions do not specify boundary markers or instructions for the agent to distinguish between its own logic and data retrieved from the environment.
Capability inventory: The skill provides high-privilege tools within the Roblox environment, including execute_luau, multi_edit, insert_asset, and input simulation tools (user_keyboard_input, user_mouse_input).
Sanitization: No sanitization or validation protocols are described for handling content ingested from scripts or console logs.
[DYNAMIC_EXECUTION]: The skill relies on generating and executing Luau scripts to interact with the game engine.
Evidence: The execute_luau tool allows the agent to run arbitrary code strings, and multi_edit enables direct modification of source files, as detailed in SKILL.md and references/full.md.
[EXTERNAL_DOWNLOADS]: The skill interacts with external registries and asset stores to retrieve content.
Evidence: Tools like search_asset, insert_asset, and search_assets in references/full.md facilitate the downloading and insertion of content from the Roblox Creator Store and inventories.

roblox-studio-mcp