better-harness

Why we included this project

Better Harness is a plugin that runs inside the coding agents you already use, like Claude Code, Codex, or Cursor, and it reviews the whole delivery loop rather than just the code that comes out. It gathers evidence from your project and sessions across five areas: task understanding, controlled execution, change validation, reliable delivery, and learning capture. From that evidence it produces prioritized findings, each one tied to visible evidence and an expected outcome, with a scoped repair plan. If something wasn't observed, it stays explicit instead of being turned into an unsupported score, and the history view shows whether the loop is actually improving over time. For engineering leads and small teams that want to tighten agent guardrails, validation, and knowledge capture without building their own harness, this is a practical way to do it.