scientific-validation

Use for changes to numerical methods, statistical assumptions, validation targets, scientific figures, or interpretation of scientific results.

SKILL.md

GitHub

Scientific Validation

  • Identify the authority source: protocol, paper, supplement, docs/SPEC.md, docs/DECISIONS.md, or tests.
  • Preserve raw/reference data and write derived outputs separately.
  • Add or update tests for invariants, reference examples, edge cases, and units.
  • Compare against known targets when available; document acceptable tolerance and why it is appropriate.
  • For figures, optimize for the comparison the reader should make and keep labels readable at manuscript size.
  • Record any assumption or divergence from the source in docs/DECISIONS.md or an ADR.