The Agent Skills Directory

[SAFE]: No prompt injection or malicious override patterns were detected; the instructions are purely focused on providing high-quality evaluation methodologies.
[SAFE]: No instances of data exfiltration, sensitive file access, or credential exposure were found within the documentation or scripts.
[SAFE]: The skill contains no obfuscated content, encoded commands, or hidden URL patterns.
[SAFE]: Code examples are for demonstration and use standard, well-known libraries for data processing and testing without performing any dangerous network or system operations.
[SAFE]: Risks of indirect prompt injection are addressed through pedagogical examples that demonstrate the use of structural delimiters for separating untrusted model outputs from system instructions.

advanced-evaluation