nx_gatescan.nx

nx_gatescan.nx -- FIND THE ORGANS NOTHING IS CHECKING. ===== WHY THIS EXISTS ============================================ Two defects were found by luck in this ecosystem within one session: nx_pow10 -- no gate of its own. Its header claimed an interpolation error of <=0.3%; measured, it is 0.66%. That figure had been inherited into a FOOD-SAFETY margin note. nx_arrhenius -- no gate AND nothing imported it. It returned exactly 1.0 for every input, i.e. "temperature does not affect reaction rate", from a file whose own comment claimed it carefully managed precision across the reciprocal step. Both were caught because someone happened to look. That is not a control. A session must remember; a tool cannot forget. This organ turns "no gate + no callers = an unverified intention" from a law into a scan. ===== THE MEASUREMENT ============================================ Composes the existing graph engine rather than rebuilding it: the whole- tree walk and accurate import edges already exist and are proven, so this adds exactly one idea on top of them. HAS_GATE -- a node named <base>_test.nx or <base>_gate.nx exists. CONSUMERS -- importers that are NOT that organ's own gate. ★THE SECOND HALF IS THE LOAD-BEARING ONE. Raw fan-in (Ca) counts the gate as an importer, so the moment you write a gate for a dead organ its Ca becomes 1 and a naive orphan check calls it "used". That is exactly what would have hidden nx_arrhenius the instant it was gated. Consumers are therefore counted with gate/test importers EXCLUDED. ===== RISK IS RANKED, NOT LISTED ================================= A flat list of ungated files invites working the easy end. The two classes are not equally dangerous and the ranking says so: UNGATED-LOAD-BEARING (no gate, N consumers) -- an error here propagates into N callers and is attributed to whichever one notices first.