QoderAI/better-harness ? reverse-engineered prompt

Build me a coding agent plugin called Better Harness that helps improve the way an agent works, not just the final code it writes.

I want it to inspect the task evidence around a session, look at things like instructions, specs, tests, hooks, and any other proof it can find, then turn that into a short report with prioritized findings and clear next steps. If something can’t be verified, keep that missing evidence obvious instead of guessing. The report should explain what the issue is, why it matters, what to fix, and how to check that the fix actually worked.

Make it work cleanly inside the coding agent experience, with host specific setup where needed, and generate a nice readable report output that I can review after a task. If there are different ways to present the report depending on the host, handle that too. Please include a simple way to run it, verify it, and show the results. Look up current docs online if you need to.