Verdicts
The intended per-submission verdict would carry:
- Classification — a Tribunal-roadmap ruleset-audit outcome (flags applied from a declared rulebook, never a judge-model quality opinion).
- Per-category pass-rate — pass-rate across the benchmark task families.
- Failure-flag list — the category/signal/severity/remediation flags raised.
- Repair-lift estimate — projected improvement after remediation.
Public discoverability of verdicts (for example via StreetLedger) is a roadmap
capability, not something available to opt into today. The built analogue that exists now
is in DefendableCloud: real verdict tiers — honey (pass) · jelly (risk) ·
propolis (fail) — recorded in per-org hash-chained receipts that
/ledger/verify can independently recompute.
🐝 Operator-grade · books and records · to the shed.