Yes so regular human and agentic evaluation of the coding agent output, scoring it on specific criteria?

https://github.com/harness/harness-evals