Agent evaluation kit
Policy checks and test cases for evaluating proposed agent work.
Independent project using synthetic cases. Not an official Nous or Hermes product, customer deployment, or accepted policy decision.
The problem
An agent completing a task is not enough to establish that the task met the organization’s requirements.
What I built
I built policy configurations, a version-specific capability map, negative tests, and an offline evaluation workflow. The final decision remains with a person.
What you can inspect
The public repository documents the offline checks and known gaps. A recorded live trial remains unaccepted; it is not presented as a completed customer result.