Umanitek's Agent Blackbox v1.2.0 added to Nous Research's Hermes plugin catalog

Umanitek says its Agent Blackbox v1.2.0 is now live in the Hermes plugin catalog from Nous Research.

The plugin checks each tool call and model request an AI agent makes against a verified threat graph on the OriginTrail DKG before it runs. Umanitek says it covers prompt injection, dangerous commands, malicious packages, unsafe skills, and attempts on wallets and keys.

According to Umanitek, the plugin can either audit these actions or block them. The source report is a post on X from Umanitek's account and gives no independent evaluation of how well the checks perform.