The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent and user to install the 'Rokit' toolchain manager by piping a remote shell script directly to a shell interpreter: curl -sSf https://raw.githubusercontent.com/rojo-rbx/rokit/main/scripts/install.sh | bash. This occurs in both the main skill instructions and the provided GitHub Actions workflow template. This method bypasses local script verification and allows a remote source to execute arbitrary code on the host system.
[EXTERNAL_DOWNLOADS]: The skill downloads type definitions and other scripts from external GitHub repositories, such as https://raw.githubusercontent.com/JohnnyMorganz/luau-lsp/main/scripts/globalTypes.d.luau, to support development tooling. These downloads introduce dependencies on third-party content that is not audited by the skill.
[COMMAND_EXECUTION]: The skill relies on the execution of multiple command-line utilities (rojo, wally, lune, stylua, selene, darklua) to manage the Roblox development lifecycle. It also demonstrates using curl to interact with external APIs, such as the Roblox Open Cloud API for publishing places.
[INDIRECT_PROMPT_INJECTION]: The skill manages and processes several configuration files which could be manipulated to influence the agent's behavior if the project source is untrusted.
Ingestion points: Reads rokit.toml, default.project.json, wally.toml, pesde.toml, and .luaurc to configure tools and project structure.
Boundary markers: No explicit markers or instructions are provided to the agent to treat the contents of these configuration files as untrusted or to ignore instructions embedded within them.
Capability inventory: The skill possesses capabilities to execute shell commands, perform network requests via curl, and write files to the local file system.
Sanitization: There is no evidence of validation or sanitization of the contents of these project files before they are utilized by the underlying tools.

roblox-tooling