nx_drift_disposition.nx

nx_drift_disposition.nx -- DECIDE + DISPOSITION the WMS dep-drift clog (the live publisher-death at scale). Operator 2026-07-04: "we need THEM to DECIDE + get these things FIXED till trusted like you are; i can't have the nishi publisher situation where the queue clogs and dies." nx_queue_health DETECTS 35 dep-drift rows (deps naming ids that don't exist -> can NEVER unblock, silently = the death) + exits RED, but nothing DECIDES them. THIS drains that RED: for each drifted row it emits a DISPOSITION worklist line (id, the phantom dep, title, gate) so the row's OWNER can decide -- REPOINT the dep to the right id, or CLEAR it if the prereq is already done/gone. Nothing stays a silent stuck row: every drift becomes a VISIBLE, owned decision. COMPOSES nx_assign_core (an_load/an_id_at/an_deps_at/an_title_at/an_gate_at/an_find) -- the SAME loader nx_queue_health uses -- so this is the DECIDE half of the existing DETECT (no fork, Cardinal 15). Writes knowledge/status/drift_worklist.txt (the owner-actionable list) + prints a summary. READ-ONLY on the queue (it does NOT auto-rewrite deps -- a mis-wired program is worse than a visible stuck row; the OWNER decides the repoint, per Claude-is-tutor-not-doer). license_tier: ORIGINAL genealogy_id: operator-2026-07-04-queue-death + nx_queue_health-drift-detector lineage_id: nishi_drift_disposition_v1 expect_exit: 0