Malware skills

[BenjaminGittins]

Hi, given the recent agent malware drama with open claw, ...

has PAI CI/CD and internal infrastructure integrated:

https://github.com/cisco-ai-defense/skill-scanner

Cheers,

• B

[danielmiessler]

No malicious skills in PAI — the project is explicitly security-focused and any skill that could be used for harm would be rejected. The Cisco skill scanner linked is interesting and worth watching as the space evolves. PAI's security model is hook-based (SecurityValidator hook) and relies on the model's own safety training plus explicit validation rules.