The Agent Skills Directory

[SAFE]: No malicious patterns, unauthorized data exfiltration, or obfuscation techniques were identified. The skill is authored by the platform vendor and follows secure resource management practices.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from the Runpod API, such as template names and resource descriptions, which could serve as an attack surface for indirect injection. However, the instructions include strict rules against echoing sensitive data and require multi-step verification, keeping the risk level at safe.
[DATA_EXPOSURE]: The skill manages sensitive resources including account secrets and SSH keys. It implements 'Hard rules' to ensure secret values are never displayed in transcripts and provides safety warnings for replacing SSH keys to prevent accidental loss of access.

lifecycle-crud