harness-engineering

Why we included this project

Harness engineering means improving what a coding agent produces by shaping the environment around it, the context and tools, rather than retraining the model. This repository is Ryan Lopopolo's field guide to that practice, gathering his theses, playbooks, and evaluation guidance into one place, with a focus on carrying an organization's nonfunctional requirements into agent context and tools. It is a curated knowledge resource, not software you install, so the value is in the structured thinking and the AGENTS.md routing file you can point an agent at alongside the system it should improve. If you have hit the limits of prompt tweaking and want the repository itself to be part of the agent's operating environment, there is concrete, opinionated material here to work from.