What problem does it solve? Drafting a research manuscript often produces prose that runs ahead of the evidence: numbers with no backing result file, conclusions stronger than the data, or negative results quietly softened. This Skill audits every claim in a draft so each one is bound to a concrete result artifact and the pre-registered decision rule that licenses it. ## Core Features & Use Cases - Claim-to-artifact binding: Links every quantitative or interpretive claim, table entry, figure caption, and limitation to a date-suffixed result JSON and vault [RESULT]/[DECISION]/[NEGATIVE] entry. - Decision-rule lock checking: Verifies that outcome-contingent prose follows the pre-registered outcome-to-prose mapping rather than a post-hoc reading. - Honest reporting enforcement: Flags overstated claims, buried nulls, and undisclosed cascade effects or pre-registration amendments, producing a claim-by-claim verdict table (TRACED / AWAITING COMPUTATION / OVERSTATED / UNDISCLOSED). - Use Case: Before submitting a paper, run the audit on the discussion section to confirm each reported effect size matches its result file and that a weaker-than-expected finding is reported honestly rather than hedged. ## Quick Start Audit the claims in my manuscript draft and verify each one traces to a result file and decision rule.
