The Agent Skills Directory

[SAFE]: The skill is composed of technical documentation and workflows for professional development use. It does not perform unauthorized network connections, attempt to access sensitive system files, or bypass security controls.
[INDIRECT_PROMPT_INJECTION]: The skill involves reading properties from the Roblox engine and console logs, which constitutes an ingestion surface for potentially untrusted data.
Ingestion points: Property values read via execute_luau in SKILL.md and runtime logs fetched via get_console_output.
Boundary markers: None explicitly defined in the prompts to separate engine data from instructions.
Capability inventory: The skill uses execute_luau, screen_capture, and start_stop_play within the Roblox Studio environment.
Sanitization: No explicit sanitization of strings read from the DataModel is described.
[DYNAMIC_EXECUTION]: The skill uses execute_luau to interact with the Roblox environment at runtime. This execution is limited to the Roblox Studio process and is the intended method for performing UI diagnostics and device simulation.

roblox-device-testing