CFO Doctrine · Full Portfolio Verifier
Meta-dashboard for all 13 synthetic cases across the 9-layer CFO doctrine map + 3 meta-skills. Fires each in order, captures sealed hash + verdict + recall, verifies UI page renders.
Layer coverage: 1 · 3 statements as system (APEX) · 2 · Frameworks (SPECTRUM) · 3 · Ratios / peers (OMNI + TRINITY) · 4 · Q-of-E forensics (AURELIA + MERIDIAN + TRINITY) · 5 · Consolidation (CASCADE) · 6 · Cross-border tax (GLOBAL) · 7 · Capital / liquidity (PILLAR) · 9 · Contagion / network (NEXUS). Plus governance (SENTINEL), macro-context (ATLAS MACRO), services P&L (SERVICE).
Cases fired
0/13
of 13 synthetic cases
Verdict · PASS
0
recall + FP + INDETERMINATE all pass
Verdict · PARTIAL
0
recall borderline
Verdict · FAIL
0
investigate immediately
Byte-determinism
—
rerun sealed hash match
Per-case scorecard
| # | Case | Layer / P | Verdict | Recall | Amount band | FP refusal | INDETERMINATE | Graph · entities / events | Brain (dry-run) | Sealed hash | UI |
|---|
Brain promote WORM ledger (KOKON→ARIN payload attempts · since service restart)
Two independent sweeps run on page load: dry-run validation, then live promotion. Counters below reflect the in-memory WORM ledger since the KOKON process last restarted; ARIN's server-side ledger holds the full history.
Dry-run sweep · contract validation
Live sweep · promotion to ARIN Brain
Read-filter sweep (AG-36 · CONTRACT_CLEAN default)
KOKON's brain client default-drops any edge whose fact_status ≠ CONTRACT_CLEAN so the r1 WITHHELD lane never surfaces in downstream reads. This strip samples a depth-1 /neighborhood on each of 13 case nodes, then reports edges kept vs dropped by the client-side filter. Drift signal: non-zero drops on a supposedly r2-only tenant.
Edges kept (CONTRACT_CLEAN)
…
Edges dropped (r1 / no status)
…
SLO status (F3 · rolling p50 / p95 / p99 per critical endpoint)
Rolling 500-sample window per endpoint. Green = within budget · red = breach · slate = under min-sample threshold (20). Live drift signal for latency degradation before an SLA breaches.
| Endpoint | Samples | p50 ms | p95 ms | p99 ms | p95 budget | Verdict |
|---|
Brain gateway status (ARIN ontology-query v1)
…probing…
Realized case outcomes (R1-R3 · hit-rate vs sealed predictions)
Every closure records realized verdict vs the sealed model prediction. Hit rate = % where realized matches predicted. Also writes CASE_OUTCOME_REALIZED to the gov ledger so the outcome enters the tamper-evident chain.
Governance ledger (G1-G3 · hash-chained · tamper-evident)
Append-only ledger for privileged operations (backups, secret rotations, chaos drills, ARIN contract runs). Each row's sealed_hash chains to the previous. The verifier walks the whole chain — any tamper breaks it. Regulator evidence surface for Wave 7.
| Seq | Timestamp | Kind | Actor | Outcome | External ref |
|---|
Brain freshness (B5 · ARIN generated_at SLI)
How stale is the ARIN-side brain data? Read from Counts.generated_at every 5 min. FRESH < 6 h · AGING < 24 h · STALE ≥ 24 h. Independent from uptime — brain can be up but serving old data.
ARIN passive health beacon (B3 · 60 s cadence · 1 h ring)
Background ticks the gateway every 60 s and stores the last 60 outcomes. Uptime % is a cheap upstream health signal without polling from the browser. Red = ARIN unreachable this window; green = live and responding.
…