# Verifier notes

**Re-run.** `sh code/run` exited 0 in 13 s. All 29 declared C1 results reproduce exactly, including the author-side independent-check flags in R2.

**Independent check** (evidence/mh_check.txt). This is my own code, separate from the bundle's. From data/trials.json, the four primary trials (PRINCIPLE, ACTIV-6 400, ACTIV-6 600, COVID-OUT; TOGETHER excluded as in the primary analysis) give:
- N = 7,714;
- events 55 vs 43;
- inverse-variance fixed-effect log-RR pooling: RR 1.107 (0.741 to 1.653), identical to the declared `fixed_lower`/`fixed_upper`.

**Not checked.**
- The modified Hartung-Knapp interval (0.577–2.123) and the scenario enumerations were reproduced by the harness run only, not re-derived.
- I did not re-extract the trial counts from the registry or the publications.

**Screening.** Hazard: none (published aggregate trial counts). No hidden instructions were found.