nx_survey_evidence_gate.nx

nx_survey_evidence_gate.nx -- the EVIDENCE-DRIVEN METRICS gate (operator 2026-07-10: "dont just make up metrics we are evidence driven use proven evidence as of 25 26 that the feedback and questions will help us identify quantitative and qualitative areas of improvement"). The nx_connect_h2h_exceed keystone applied to METRIC SELECTION: a shipped analytic is EVIDENCED only when (a) its implementation symbol exists in runtime/nx_survey_stats.nx on disk, AND (b) a VERBATIM needle from the corresponding banked 2024-2026 corpus (knowledge/library/swc_svy_*.txt, fetched keyless from OpenAlex over sovereign TLS) is found in the banked bytes -- real literature tying that metric/question class to identifying improvement areas. Three verdict classes per row: EVIDENCED symbol present + needle present (we ship it, the field backs it) EVIDENCED-GAP symbol ABSENT + needle present (the field says it matters; our honest build list -- the evidence itself ranks the roadmap: attention checks, LLM text coding) REJECTED needle absent (fabricated citation -> the row DIES; the neg-control proves this fires) Exits 0 iff every real row lands its declared class AND the fabricated row is REJECTED. license_tier: ORIGINAL expect_exit: 0