The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses clear, strict instructional language to guide the agent through a step-by-step workflow. There are no attempts to bypass safety filters, extract system prompts, or override core agent instructions.
[DATA_EXPOSURE]: The skill does not access sensitive local file paths (such as SSH keys or environment variables) or hardcode secrets. All required credentials, such as App Keys and Ad Unit IDs, are handled through user-supplied placeholders within C# code templates.
[EXTERNAL_DOWNLOADS]: The skill references the official 'Ads Mediation' package (com.unity.services.levelplay) via the Unity Package Manager and standard Android/iOS dependency managers. These are trusted vendor resources from Unity Technologies.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where it ingests user-supplied configuration strings (App Keys and IDs) to populate C# scripts. However, this is the primary intended function of the skill, and the output consists of static source code files intended for manual review and saving by the user, which serves as a natural security boundary.
[DYNAMIC_EXECUTION]: No dynamic execution of code or load-time shell commands (!command) was detected. The C# scripts generated are for the developer to integrate into their Unity project and are not executed by the agent itself.
[OBFUSCATION]: No hidden content, Base64-encoded commands, homoglyph attacks, or zero-width characters were found in the skill's text or instructions.

levelplay-unity-integration