The Agent Skills Directory

[Prompt Injection] (SAFE): No patterns detected that attempt to bypass safety filters or override core system instructions. The instructions are focused solely on metadata formatting and correctness.
[Data Exposure & Exfiltration] (SAFE): No hardcoded credentials, sensitive file path references, or network commands were found in the skill content.
[Remote Code Execution] (SAFE): The skill does not download external scripts or use dynamic execution functions like eval() or exec().
[Persistence & Privilege Escalation] (SAFE): No commands related to system persistence, sudo usage, or permission modification are present.
[Indirect Prompt Injection] (SAFE): Although the skill accepts file input for review, it does not possess capabilities such as file writing or command execution that could be exploited through malicious content in the processed files.

fixing-metadata