kouyichi/dsh-bench

“Agent benchmark runner for dsh: run a task set through fresh agents, score pass/fail by expectation keywords, measure duration, and produce comparison reports (evaluation gap in the ecosystem).”