The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for handling untrusted data from the game environment, such as player IDs and game state tables, which could potentially contain malicious content if later processed by an agent.
Ingestion points: External data enters the context through player.UserId and gatherDataFor(player) functions in SKILL.md.
Boundary markers: No specific boundary markers or "ignore instructions" warnings are applied to the data values themselves.
Capability inventory: The skill includes database read and write operations via DataStoreService:GetAsync and DataStoreService:UpdateAsync in SKILL.md.
Sanitization: Explicit data validation or sanitization of the saved values is not provided in the examples.
[EXTERNAL_DOWNLOADS]: Includes a reference to official technical documentation on the roblox.com domain, which is a well-known service, for details on error codes and platform limits.
[SAFE]: The skill promotes several security best practices, such as using pcall for network-bound operations to prevent execution errors and advising users to keep Open Cloud API keys secret.

roblox-datastores