nx_explore_census.nx

nx_explore_census.nx -- THE NISHI EXPLORE TEAM: capability census + living map (operator 2026-06-23: "build a nishi explore team capability from the hardware rung up each rung to be s class exceed"). WHY THIS EXISTS (the dogfood): this session the third-party Explore *agent* hit a spend wall AND ripgrep-backed Glob TIMED OUT FOUR TIMES at 20s on this very tree. The sovereign explore capability is justified by DIRECT MEASUREMENT, not assertion -- the incumbents demonstrably fail here. This organ is the ladder's RUNG-0 keystone: it does NOT re-implement anything. It PROBES the real on-disk organs that already provide each explore rung and grades them HAVE/MISSING by reading the file for its DEFINING SYMBOL (so it cannot lie -- drop the symbol and the rung flips MISSING). It is SELF-GATING: baked neg-controls (a bogus-symbol liar-kill + an absent-file open-fail control) must hold or it exits RED. Composes nx_codegrep's proven read+substring idiom. license_tier: ORIGINAL THE LADDER (each rung: incumbent / sovereign exceed axis / proving organ): R0 WALK recursive fs-walk via getdents64 vs find/os.walk nx_dir.nx::nx_dir_list (JPL-bounded, verdict-sealed, no libc -- bounded where find recurses unbounded) R1 GLOB name/suffix match vs find -name nx_dir.nx::nx_dir_name_ends_with R2 GREP recursive content search vs ripgrep nx_codegrep.nx::cg_scan_dir (COMPLETES where rg TIMED OUT on runtime/ -- measured live, 4x this session) R3 INDEX durable inverted index + BM25 rank vs Sourcegraph nx_library_harvest_v2.nx::lh2_query R4 MAP query -> conclusion map (paths+role+hits+rank) = the EXPLORE-AGENT REPLACEMENT, the GAP R5 GROW self-discovers its own coverage gaps and grows the GAP HAVE rungs are proven on-disk; MISSING rungs are the HONEST backlog (the gap IS the roadmap). Build/run: ./_offc/nx_sov_build_run.elf nx_explore_census (run-exit=0 => map correct + neg-controls held)