Python 3; standard library only. No model access or infrastructure credentials required.
How to use it
Download both files into one directory. Run python3 incident-agent-grader.py --self-test, then python3 incident-agent-grader.py sample-observation.json. Read the sample observations beside the verdict so you can see which evidence produced it.
Scope
The tests check grading logic only. The sample does not establish production recovery or benchmark an AI model.
