The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions and technical references describe a workflow where the agent reads data from the Roblox Studio environment to inform its actions, creating a surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through tools like script_read, get_console_output, script_search, and search_game_tree as defined in references/full.md.
Boundary markers: The instructions do not specify the use of boundary markers or delimiters to isolate ingested content, nor do they warn the agent to ignore instructions embedded within game scripts or logs.
Capability inventory: The agent possesses high-impact capabilities including execute_luau (arbitrary code execution), multi_edit (modifying scripts), and insert_asset (downloading and inserting external assets).
Sanitization: There is no mention of sanitization, filtering, or validation of the content retrieved from the Studio session before it is processed by the LLM.
[DYNAMIC_EXECUTION]: The skill facilitates the generation and execution of code and assets at runtime within the Roblox environment.
Patterns: The execute_luau tool allows the agent to run arbitrary code segments, and insert_asset enables the dynamic loading of content from the Roblox Creator Store or inventory into the active game session.
[COMMAND_EXECUTION]: The skill provides a comprehensive interface for controlling the Roblox Studio application through various commands.
Patterns: Tools such as multi_edit, start_stop_play, and character_navigation allow the agent to execute complex operations that modify the workspace and simulate user interactions.

roblox-studio-mcp