nx_genealogist_gapscan.nx

nx_genealogist_gapscan.nx -- the Nishi GENEALOGIST's gap-scanner (FamilySearch-for-code: "find the missing gaps"), sovereign NishiLang (raw syscalls, no shell/sed/awk). (renamed from nx_genealogist to un-shadow the DISTINCT _hdl_build/nx_genealogist anti-DUPLICATION library -- that one prevents rebuilding what exists; this one finds what is MISSING. Two facets, one old name.) Mechanizes the Genealogist leg of the "what comes next" plan: it scans a registry of modules for GAP MARKERS (TODO_SILICON = an unbuilt/stub silicon path; SEED = a not-yet-matured module) and reports which modules carry which. Each gap it finds is fused (in NISHI_WHAT_COMES_NEXT.md) with the RESEARCHER's target for that capability (NISHI_RESEARCH_BASIS.md) to produce a ranked next-step -- so the roadmap is DERIVED from code-state x research, not hand-curated, and re-derives itself as we build (gaps shrink, plan updates). DATA-DRIVEN (Cardinal 11, fixed 2026-06-16): the scan targets and the DRY threshold are NO LONGER hardcoded. The old organ baked `let N=4` / `let M=6` + two arrays of absolute paths straight into code -- the exact magic-number / data-in-code smell. Now the target list lives in DATA (knowledge/registry/genealogist_targets.tsv: kind<TAB>name<TAB>path), the counts are DERIVED from the row count, and the DRY threshold is one named const. Add/remove a module = edit the data file, no code change (Cardinal 25); the scan re-derives from current data every run (fresh, not a baked snapshot). Bounded length-safe substring scan over the raw file bytes (sys_read_file), so it never over-reads. license_tier: ORIGINAL