What problem does it solve? Manuscripts often contain numbers that drift out of sync with the analysis outputs that produced them, and manual checking before submission is slow and error-prone. This Skill systematically compares every numeric claim in a paper against the actual computed results and records the verdict in a reproducibility passport. ## Core Features & Use Cases - Claim extraction and matching: Pulls point estimates, standard errors, p-values, sample sizes, and percentages from a manuscript and matches them to values in .rds, .tex, .csv, .parquet, .dat, .log, and .json outputs with a confidence score. - Tolerance-based dispositions: Applies default or per-claim tolerances and assigns PASS, FAIL, EXPLAINED, or UNMATCHED, with strict rules preventing vague explanations from downgrading failures. - Passport lifecycle tracking: Creates or refreshes a YAML reproducibility passport at quality_reports/passports/<paper-slug>.yaml, tracking STALE and UNVERIFIED states and a conformance ladder from index to verify. - Use Case: Before submitting a revised paper, run the audit against the manuscript and the results directory to confirm every reported statistic still matches the current analysis outputs, then update the passport with fresh verification timestamps. ## Quick Start Audit the numeric claims in manuscript.qmd against the outputs directory and update the reproducibility passport for this paper.
