KB-5DCC

T2 RP Count Reliability Audit — 00 Readme First

3 min read Revision 1

Terminal 2 — Registries-Pivot Count Reliability & Bug Audit (2026-06-05)

Auditor: Terminal 2 (parallel, READ-ONLY). Scope: Whole Registries-Pivot count chain — base pivots, AX-TOPIC, AX-PROCESS, AX-TRIGGER, Process×Trigger, DOT/tools, aggregate pivots, final substrate, adapter coverage, UI contract. Core question: Are the things Registries-Pivot counts actually counted correctly, from trustworthy sources, without duplicate/missing/stale/hardcoded/blind-spot errors?

Read-only confirmation

NO mutation to DB, filesystem, UI, Git, approvals, owners, events, REAL_RUN, source IU, or KB source content. Only audit docs under this report path + the checkpoint were written. All DB access via query_pg (read-only role, READ ONLY txn, 5s timeout, LIMIT 500) against the directus database.

Where the system actually lives (corrects a common assumption)

The Registries-Pivot system is in the directus Postgres database (public schema, 225 pivot/axis/wf/dot objects), NOT in workflow or incomex_metadata. query_pg allowed DBs: directus, incomex_metadata, workflow. dot_config is a native BASE TABLE in directus and a FOREIGN (FDW) table in incomex_metadata.

Headline verdict

RP_COUNTS_RELIABLE_AND_HONESTLY_LABELED_FOR_DB_NATIVE_DATA; PARTIAL_FOR_HOST_AND_KB_LAYERS.

  • Every internal count chain audited reconciles exactly (topic 7→25, process 19→143, trigger 7→525, drill children, aggregate pivots → live source).
  • Counts are LIVE-computed in the views (verified: PIV-019 = live birth_registry 1,194,844; PIV-007 = live dot_tools 309; PIV-311 = live information_unit 219; PIV-313 = live system_issues(open) 220,170).
  • No TRUE_BUG producing a silently-wrong count was found. Every divergence is surfaced via a warning_flags / drift_status / detector_status / coverage_status column.
  • Real weaknesses are adapter blind spots (host crontab/systemd/fs scripts/KB SOP), DOT registry↔filesystem reconciliation, and the fact that nothing is officialized (all CANDIDATE — correct, owner/president-blocked, not a bug).

Overall reliability score: 76 / 100

Area Score
Base RP layer 83
AX-TOPIC 80
AX-PROCESS 74
AX-TRIGGER 78
Process×Trigger 80
DOT/tool inventory 68
IU/document/system-issue pivots 82
Final substrate 80
Adapter/source coverage 62
UI contract readiness 80
Overall 76

Document map

  • 01 base RP count reliability
  • 02 topic axis reliability
  • 03 process axis reliability
  • 04 trigger axis reliability
  • 05 process×trigger reliability
  • 06 dot/tool count reliability
  • 07 aggregate pivots reliability
  • 08 adapter/source coverage
  • 09 ui contract risk
  • 10 bugs/gaps and priorities
  • 11 final summary

All numbers in these docs are LIVE as of 2026-06-05 ~02:36 UTC.

Back to Knowledge Hub knowledge/dev/reports/architecture/parallel-terminal2-registries-pivot-count-reliability-bug-audit-2026-06-05/00-readme-first.md