The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and analyze existing scripts, modules, remotes, and project mapping files (SKILL.md, Step 2). These represent ingestion points for untrusted data that could contain malicious instructions targeting the AI agent.
Ingestion points: search_game_tree, script_search, script_grep, and script_read tools access project files and metadata.
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat content within these files as data rather than instructions.
Capability inventory: The agent has access to execute_luau (code execution), multi_edit (file writing), and insert_asset (external content ingestion).
Sanitization: No sanitization or validation of the content read from project files is specified before it influences the agent's planning or execution.
[DYNAMIC_EXECUTION]: The skill utilizes the execute_luau tool to run code directly within the Roblox Studio environment. While this is a core requirement for testing and verification, it represents a high-privilege dynamic execution capability.
[COMMAND_EXECUTION]: The skill employs powerful tools such as multi_edit for modifying project source files and insert_asset for downloading and inserting external models and scripts into the project. The skill author correctly notes that inserted models can contain third-party code (Scripts) and advises inspection.

roblox-studio-workflow