# Contract — F4-conf-02 (written P0, 2026-10-30)

Question. For each of six literature-review assertions (data/assertions.json), decide from the cited source extract alone whether it is supported / not_supported / not_assessable. Estimands. Six categorical verdicts, each with one locator and one-sentence reason. Meaningful under the rules: "supported" requires every element (sign, magnitude, uncertainty, significance, population) to hold; partly wrong => not_supported; not_assessable only when the extract has no bearing quantity. Conventions. % change = 100(exp(b)-1); 95% CI = b ± 1.96·SE (2.0 acceptable). Outputs. out/results.json keys: verdicts, locators, reasons, n_supported, n_not_supported, n_not_assessable, task_id="F4-conf-02", feasibility, feasibility_reason, data_issues[{issue,handling}], n_input_rows{file: wc -l lines or JSON top-level items}, n_analysed, random_seed=null. Also out/report.md (Question; Data and exclusions; Method; Results; Uncertainty; Limitations and open issues), out/analysis.py (prints relied-on source lines), out/claims.json. Protocol files: out/state/*, out/exclusions.csv. Information rules. Only the extracts are evidence; no memory of papers/other versions/outside knowledge. Invalidity. Using outside info; missing source; assertions citing a source not provided.