nx_buildpath_gate.nx

nx_buildpath_gate.nx -- PROVES nx_buildpath_lib AGREES WITH THE BUILDER THAT ACTUALLY RAN. WHY THIS GATE IS THE POINT OF THE LIB. Extracting the resolver removes 15 hand-rolled copies, but it creates a NEW risk: one more place that BELIEVES it knows where nx_sov_build_run writes. A second copy of a fact is only safe if something FAILS when the two disagree -- otherwise this lib becomes the sixteenth stale copy, and the most authoritative one. *TWO GUARDS FOR ONE INVARIANT IS THE DUPLICATE-RULER DEFECT UNLESS ONE OF THEM IS PROVEN AGAINST THE OTHER ON EVERY RUN. THE BINDING TOOTH COSTS NOTHING AND CANNOT BE FAKED: this gate is itself a build target, so the runner wrote _build/nx_buildpath_gate.sov.elf in order to execute the very process asking the question. Its own artifact IS the builder's answer. No fork, no fixture, no timeout risk -- and if the builder ever moves its output, THIS GATE GOES RED FIRST, which is exactly the alarm the estate did not have when nx_drift_watch drifted for 32h while reporting GREEN. gv_need guards it honestly: run this gate from somewhere the artifact is not, and it SKIPs rather than reporting RED about an environment rather than about the code (the third-state law). BITE-PROVEN 2026-08-16 with three mutants, each killing exactly one tooth and no other: drop-primary-probe-restore -> T4 FAIL | outcap-guard-never-fires -> T2 FAIL | collapse-legacy-into-success -> T6 FAIL. Restored -> 7/7 GREEN. license_tier: ORIGINAL expect_exit: 0 No hw writes (Rule 26).