What problem does it solve? Empirical analyses often produce numbers that run cleanly but are silently wrong — stale figures, unreconciled totals, or results that only exist due to cached session state. This Skill enforces evidence-before-assertion checks so a result is verified, not just submitted, before it ships. ## Core Features & Use Cases - Reconciliation and clean-room reproduction: Ties totals to source by an independent path and re-runs the analysis from raw inputs in a fresh kernel with a fixed seed, in R, Julia, Python, or Stata. - Reliability vs. validity checks: Confirms determinism and approved robustness, then anchors the number's level against external benchmarks, known shocks, or alternative construct definitions. - Economic interpretation pass: Converts coefficients into interpretable units, checks economic significance, magnitude plausibility, mechanism consistency, and literature benchmarks. - Use Case: Before reporting a regression estimate, ask the assistant to verify it — it reconciles the headline number to source, reproduces it from a clean session, confirms the approved robustness ran, and ties every table cell to the current run's output. ## Quick Start Ask the assistant to verify this analysis result by reconciling the totals to source and reproducing it from a clean session before reporting. ## Core Features & Use Cases See above for the full checklist-driven verification workflow. ## Quick Start Ask the assistant to verify this result before shipping it.
