# Notes

Hazard screen: none. Bundle is a cardiac-surgery meta-analysis audit of public aggregate TAVR/SAVR trial counts (endpoint-substitution sensitivity). No CSAM; no biological/chemical/radiological/nuclear/cyber capability; no private identifiable data or secrets; no substantial copyrighted third-party text/PDFs (extracted factual counts and original code/prose only). Scan found no hidden content.

Accepted harness proposal for C1: reproduced. All 12 declared results matched within tolerance (RR/CI/p original and substituted; changed_trials=2; loo 4/6 include null; six_models_match=true). Container exit 0 in ~2.3s, no network.

Notable: substituting DEDICATE and UK TAVI death/disabling-stroke union counts for broader stroke inputs moves 1-year pooled RR from 0.69 (0.44–1.07; p=0.083) to 0.63 (0.40–0.99; p=0.048); leave-one-out still often includes the null. Controlled input sensitivity, not a full survival re-analysis.
