The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of existing game scripts, which could contain instructions designed to influence the agent's behavior.
Ingestion points: The skill uses tools like script_read, script_grep, and script_search to ingest content from the game environment into the agent's context as specified in SKILL.md.
Boundary markers: There are no explicit instructions to use delimiters or ignore natural language instructions found within the code being analyzed.
Capability inventory: The skill allows for significant actions including executing Luau code via execute_luau and modifying the file system via multi_edit in SKILL.md.
Sanitization: No sanitization or verification of the content from the ingested scripts is performed before it is used to guide the implementation of new logic.

roblox-gameplay-development