Second_Loop_Engineering

In agent-run analytics and engineering workflows, outputs must cite their evidence; when they fail, teams tune prompts and checks by hand without a fixed measure. This project automates that loop for any such workflow. Three agents (data engineer, statistician, lead engineer) form an agentic engineering team inside a harness that learns from its recorded runs. Each improvement cycle finds root causes and proposes changes to its checks, stages, and what each agent sees. A fixed evaluator and five acceptance rules, set before any code, decide what stays. MongoDB Atlas is its memory: events, proposal history, and lessons retrieved by Atlas Vector Search.