nx_matrix_sym_lib.nx

nx_matrix_sym_lib.nx -- THE ONE READER OF A COMPARE-MATRIX ROW'S SYMBOL FIELD (field 3). WHY THIS LIB EXISTS (measured 2026-08-16). The matrix format documents THREE symbol forms: <real_symbol> the row claims a capability, grounded by finding it in the organ _ABSENT_ a declared gap, owning nothing _ABSENT_:<future_symbol> a WATCH CONTRACT -- the compare hive's build order. The generator measures the named symbol every emit and flips the published cell when the workstream ships it. FIVE separate organs each re-derived that format privately, and EVERY ONE of them knew only the first two forms, so every watch row was misread -- each in its own direction: nx_sota_status `f2l == 8` -> watch row counted as a DELIVERED CAPABILITY. mangagen published cov=1000 gaps=0 with SEVEN unbuilt capabilities. An OVER-CLAIM on the one number whose entire purpose is to refuse over-claiming. nx_domain_admit grounded the literal "_ABSENT_:mp_page_emit" in the organ source -- a string that can never occur -- so every watch row FAILED ITS DOMAIN'S ADMISSION. nx_capgraph_edges watch rows were treated as OWNING their organ, inventing ownership edges. nx_compare_unified / nx_swarm_maturity same exact-equality test, same blind spot. A FORMAT KNOWN IN FIVE PLACES IS FIVE FORMATS, AND THEY DRIFT APART SILENTLY BECAUSE EACH COPY LOOKS CORRECT ON ITS OWN. The tell was that the SAME suffix broke all five the same day it appeared. SCOPE, DELIBERATELY NARROW: this lib owns the FORMAT (what the forms are, where the prefix ends). It does NOT own POLICY, because policy legitimately differs and collapsing it would be a second defect: a purely lexical consumer must treat an unlanded watch row as a gap, while a consumer that already reads the organ should ground the REAL symbol and count a landed contract as delivered. Both are correct; they answer different questions. So the lib returns FACTS, never verdicts. NO FILE IO HERE ON PURPOSE -- that is what makes it usable by the lexical consumers too. license_tier: ORIGINAL