cryer / agent-eval

git clone https://github.com/cryer/agent-eval.git

CI for LLM agents — change the prompt, swap the model, edit a tool, refactor the loop: replay the eval suite and let statistics decide if anything got worse. OTel tracing, calibrated LLM-as-judge,adversarial simulated users, and a GitHub Actions gate that blocks significant regressions.