The Agent Skills Directory

[SAFE]: The skill serves as a technical manual and code repository for VRChat developers. It contains no executable code outside of standard UdonSharp (C#) templates which follow security best practices, such as prefixing public methods with underscores to block legacy network dispatch vulnerabilities.
[DATA_EXPOSURE_&_EXFILTRATION]: No sensitive data, API keys, or private credentials were found. All identifiers used (such as World IDs like wrld_xxxxxxxx...) are public placeholders or official examples. There are no unauthorized network operations; all referenced URLs point to official VRChat domains (e.g., creators.vrchat.com, ask.vrchat.com) or the author's own development repositories.
[PROMPT_INJECTION]: No prompt injection patterns, bypass instructions, or DAN-style overrides were detected. The language is professional and instructional, focusing on Unity and VRChat development.
[OBFUSCATION]: No hidden content, base64-encoded commands, or malicious homoglyphs were found. Placeholder text in non-Latin scripts (e.g., in documentation examples) is clearly contextual and not used for deceptive purposes.
[REMOTE_CODE_EXECUTION]: The skill does not perform any external package installations or remote script executions. Dependencies mentioned are standard VRChat SDK packages which are considered trusted within the specified development environment.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external data (such as SDK error messages provided by the user), it lacks the capabilities (like network writing or sensitive file access) that would make such an injection exploitable. The risk is assessed as safe.

unity-vrc-world-sdk-3