experiment 2
DSH 插件:同一任务、同一代码基线,多 Variant(模型/Preset)并行对照实验、EvidenceReceipt 对比与胜出 Patch 导出
dsh plugin add dsh-bakeoffDSH Agent experiment & A/B comparison — run the same coding task in two isolated workspaces and compare objective metrics (wall time, turns, tokens, diff, evaluator results).
dsh plugin add dsh-run-lab