The Agent Skills Directory

[SAFE]: The skill provides documentation and architectural guidelines for integrating with the Tiny World Builder platform. No malicious patterns or security vulnerabilities were detected.
[CREDENTIALS_UNSAFE]: References sensitive environment variables (e.g., LIVEKIT_API_SECRET, TINYWORLD_AUTH_SECRET) as part of the configuration requirements but does not expose actual secrets or hardcoded keys. It correctly advises the use of environment variables for secrets.
[INDIRECT_PROMPT_INJECTION]: Identifies surfaces for untrusted user input, such as community profiles and world payloads. The instructions mandate the use of a two-layer safety filter (checkTextSafety) and specific normalization procedures for external data.
[DATA_EXPOSURE_AND_EXFILTRATION]: Contains instructions to prevent data exposure by mandating that identity verification targets must not be derived from request-controlled origins and ensuring profile assets are restricted to allowlisted presets.

tinyworld-integrations