00 — README First — T2 RP Systemic Reality / Hardcode / Auto-Scale / Design-Gap Audit

Date: 2026-06-05 · Mode: READ-ONLY (Terminal 2 Auditor) · DB: directus (RP lives in directus.public, NOT incomex) Verdict: PASS — audit useful, all known/visible systemic gaps exposed, every uncertain area labeled honestly.

What this audit is

A brutally-honest architecture-reality audit of Registries-Pivot (RP) as a supervision UI that must: reflect real PG/system structure, drill correctly, count reliably, and auto-scale without hardcode. It does NOT re-verify counts for their own sake — it audits whether the architecture is trustworthy and self-extending.

Read-only confirmation

• Mutations: NONE except this KB audit doc set (12 docs + 1 checkpoint under this report dir).
• No DB/FS/UI/Git/approval/owner/event/REAL_RUN/IU/KB-source mutation.
• All evidence gathered via query_pg (read-only role, AST-validated, 5s timeout) + pg_get_viewdef + KB doc reads. UI Nuxt source (/opt/incomex/docker/nuxt-repo/web) is outside the read allowlist → marked UNVERIFIED_SOURCE_ACCESS; UI audited from package docs only.

Headline scores (0–100)

Dimension	Score
Hardcode safety	52
Auto-scale readiness	38
Source adapter coverage	62
Count reliability	70
Drill full-population readiness	45
Birth/governance synchronization	58
Automation reality	40
UI architecture readiness	55 (design) / UNVERIFIED (source)
Overall RP trustworthiness	54

The one-paragraph truth

RP is honest, birth-safe, and no-fake-green — every divergent count is column-labeled, REAL_RUN is fail-closed, 0 nodes are faked official. But it is not yet a trustworthy supervision UI: (1) the central v_rp_universal_node_ui_contract is a 6-way UNION ALL of hardcoded axis literals — it does not read axis_registry, so a new axis requires editing the view; (2) there is a dual-contract divergence — the v1 contract still live-reports AX-PXT 10 NEEDS_GROUPING + AX-PROCESS 2 substrate_missing while only the non-deployed v2 twin resolves them; (3) the scanner orchestrator has not fired on 06-05 (wf_scanner_run_log last run 2026-06-04 09:53); (4) AX-PXT counts are static ledger literals, at least one stale (PROC:residual_reconcile=8 vs live 2); (5) the proof matrix is curated, not full-population, and its blocked branch hardcodes a literal PASS_WITH_EXPECTED_BLOCKER verdict that can never FAIL; (6) 0 of 453 process nodes are official, 0 ownership, 5/5 PROC-OWN pending — by design, authority-gated.

Document map

• 01-hardcode-audit.md — every hardcode, classified
• 02-autoscale-readiness.md — what happens when a new axis/source/object is added
• 03-source-adapter-reality.md — adapters: live/stale/partial, last scan
• 04-count-reliability-systemic.md — silently-wrong vs misleading vs stale counts
• 05-full-population-drill-risk.md — curated-proof risk, route-to-empty risk
• 06-birth-governance-rp-sync.md — synthetic-vs-registered, candidate-vs-official
• 07-automation-reality.md — what actually fires vs plan-only
• 08-ui-architecture-audit.md — Nuxt math, hardcoded routes, labeling (package-doc level)
• 09-design-gap-audit.md — additions vs intended architecture
• 10-fix-roadmap-priorities.md — P0/P1/P2/P3, root cause, who fixes
• 11-final-summary.md — consolidated verdict + scores

Relationship to T1

This audit does not contradict T1's engineering claims — grouping/proof/substrate were genuinely built and the v2 views genuinely resolve them. It contradicts only the framing that the system is "production-closeout ready": the resolved state lives in non-deployed v2 twins, the deployed-intended v1 contract is stale, automation is unproven, and the contract is not registry-driven. T1's "NO engineering blocker" is true for authority; this audit surfaces architecture debt that is an engineering concern T1 can fix without authority.