Enterprise evidence · honesty boundary
What SinkAlert does not claim
One page for governors, OIC judges, and GISTDA reviewers. Prefer Contabo (dashboard.cfoth.ai). Phase B RF stays research-only.
Hard line: no fabricated predictive risk scores · no auto-CAP · no priced insurance premiums · no “risk right now” authority from research RF.
We do NOT claim
- Citywide operational early-warning accuracy or deployable susceptibility for BMA dispatch.
- Automatic CAP / public warning issuance (
cap_auto_issue=false forever until agency authority + SOP).
- That denser catalog labels (54 BMR / 49 unique) beat frozen web
phase_b_rf_v1_56 — held retrains stay held.
- OOF AP / F1 / LOPO AUC as citywide KPIs or insurance rating factors.
- afolio.co parity with Contabo until Hermes wrangler redeploy is curl-verified.
- Demo Basic Auth (
admin:555) as government SSO.
We do claim (verifiable)
- Verified news incident catalog + human-reviewed citizen / LINE intake on Contabo.
- Live MHEWS JSON:
/api/mhews/health and /api/mhews/phase-b/meta with operational_use=false.
- CAP-shaped coordination drafts with human review/approve flags — drafts ≠ issuance.
- Frozen open-data RF screening layer for pilot research; hard FN class (e.g. Sai Noi) documented.
- OIC Challenge #3 decision-support narrative — screening & ops, not actuarial pricing.
Still blocked (external)
- Partner underground GIS (MEA / MWA) and numeric PSI — see Partner GIS ask.
- SSO / IdP for district desks.
- Agency CAP issuance connector + legal authority.
- Cloudflare afolio redeploy (docs/105).
Canonical: docs/000-AGENT-MAIN-REFERENCE.md · Freeze: docs/100-phase-b-rf-v156-FREEZE.md · Adversarial: docs/113, docs/114. Do not brief RF OOF as operational accuracy.