A
licenseNot graded
qualityC
maintenanceEnables running an agentic multiverse analysis of a published empirical paper by driving fresh Claude Code or Codex agent sessions through package discovery, reproduction of the headline estimate, a pre-registered grid of 100–250 defensible specifications, and independent auditing. It reports how much of the paper's defensible specification space preserves its main conclusion, exposing runs as status, results, and file-reading tools plus pipeline documentation and role prompts.
MIT