11 — Final Summary

Verdict: PASS (audit complete; all uncertain counts labeled honestly)

Read-only. Mutations: NONE except these KB audit docs.

Overall Registries-Pivot count reliability: 76 / 100

Area	Score	One-line
Base RP layer	83	Live counts, clean parent graph, 12/12 drill proof; 14 scaffold nodes no-count, labeled
AX-TOPIC	80	7 nodes = 25 assignments exact; all candidate/orphan, unratified source
AX-PROCESS	74	19 candidates = 143 members exact; headline mixes noise/infra; 2 official workflows axis-invisible
AX-TRIGGER	78	7 nodes = 525 exact; detector distinguishes stale/inactive; host triggers excluded
Process×Trigger	80	12-row ledger fully attributed (severity+owner+action); all relations candidate
DOT/tool inventory	68	309 clean registry but only 119 script_path / 41 no-file; true schedule off-DB
IU/doc/system-issue pivots	82	PIV-311=219, PIV-313=220,170 exact to live; PIV-301/302/303/310 don't exist
Final substrate	80	85/87 available; 2 empty placeholders; substrate≠count semantics
Adapter/source coverage	62	DB native & live; host ingest-only partial map; KB unknown/FP-risk
UI contract readiness	80	Complete (0 null routes), server-computed, no Nuxt math; 3 synthetic axes hardcoded

The one sentence

Registries-Pivot counts the DB-native world correctly, live, and labels every uncertainty — but it cannot yet be trusted for the host/KB world (cron, systemd, scripts, SOP docs), and nothing is officialized, so a UI can show DB counts now with candidate/partial warning labels but must not present host/KB-derived or "official" counts as complete.

Critical bugs

NONE (no silently-wrong count). Closest issues are stale parallels (trigger_registry 107), test noise (MTX-TEST), and phantom references (PIV-301/302/303/310 don't exist) — all low/contained.

Count-unreliable areas

AX-PROCESS "19 candidates" (lane-mixed); AX-TRIGGER "52 event types" (22 inactive) & "525" (DB-only); dot_tools 309 as executables; KB SOP count (unknown).

Adapter blind spots

KB SOP enumeration (count unknown, FP-risk) · host_crontab (true schedule, DB-invisible) · systemd_timers · fs_scripts · host triggers excluded from trigger universe.

Candidate / owner-blocked but NOT bug

official AX-PROCESS=0; both registered axes CANDIDATE; 0 official nodes; 12/12 ledger gaps safe_now=false; job:cut READY_FOR_PRESIDENT; real_run_enabled=false, REAL_RUN=0 (6 obs all DRY/SIM).

Must fix before trusting UI counts

1. Label all host/KB counts partial/unknown. 2. Show AX-PROCESS by lane, AX-TRIGGER by active/inactive, trigger universe as "DB/dot only". 3. Reconcile dot_tools↔bin/script_path or relabel. 4. Remove phantom PIV-30x/310. 5. Retire MTX-TEST; deprecate stale trigger_registry.

Can be shown now (with warning labels)

Base RP live counts (grey no-count nodes); AX-TOPIC (candidate badge); job:cut + 6 genuine process candidates (candidate badge, noise segregated); AX-TRIGGER 525 ("DB/dot" + active/inactive); Process×Trigger ledger (worklist); PIV-311 & PIV-313 (trustworthy, live).

Mismatch with T1 assumptions

If T1 (the dynamic-drill UI/PG-structure proof) asserts "UI_DYNAMIC_DRILL_PROVEN across 5 axes / 87 nodes / 12-12 PASS" — T2 CONFIRMS that at the contract/PG level (verified live: 87 nodes, 12/12, 85/87 substrate, 0 null routes). T2 adds the caveats T1's proof scope does not cover: (a) only 2 of 5 axes are governance-registered (3 synthetic); (b) the 12-case proof is curated, not full-population; (c) host/KB adapter coverage is partial; (d) substrate_available=true does not imply a non-empty/known count. No contradiction on the numbers T1 proved; T2 widens the reliability envelope around them.