Lewati ke isi

T0-E1.1 independent verdict

Verdict

E11-C — T0-E1.1 FAILED — MATERIAL SECURITY OR EVIDENCE FAILURE REMAINS

The dominant blocker is S0 false-green evidence: the frozen manifest does not describe the Golden runtime that produced the claimed results, and the validator returns PASS despite that mismatch. T0F-006 also retains a production-capable predictable-password provisioning default. T0F-004, T0F-005, T0F-007, T0F-008, T0F-009, T0F-014, and T0F-015 are not independently closed because their evidence is incomplete or not bound to a trustworthy runtime.

T0F-002 and T0F-013 remain closed by regression review only. T0F-010, T0F-011, and T0F-012 remain outside scope and open for T0-E2. Full T0 closure is not declared.

Final adversarial answers

  1. Not proven closed. Source is improved, but runtime conflict evidence is not trustworthy; remaining group/legacy fallbacks require further review.
  2. Not proven closed. The sensitive-model gate reports denial, but alternate methods/relations are not fully evidenced on the exact runtime.
  3. Not proven closed. Must-change dataset denial is reported, but evidence binding fails.
  4. Not proven closed. Cross-tenant reset denial is reported only by mismatched-runtime evidence.
  5. Yes. provision_clean_tenant.sh still has a default ADMIN_PASSWORD=admin path and prints/uses it.
  6. Not proven. Invalid-token signup is blocked; valid-token establishment was not independently tested.
  7. Not proven by independent runtime. The matrix claims one open token, but its concurrency evidence is not independently bound.
  8. Not proven. Code claims fail-closed; no independently bound failure injection.
  9. Not proven. Same audit limitation.
  10. Not proven. Negative API probes exist, but positive entitlement and native RPC boundaries are incomplete.
  11. No independent positive proof. The required enable-one-package test is absent.
  12. No. Fixture checksum is null, module list is empty, and entitlement profile is empty despite live non-empty state.
  13. No. Backend identity trusts an environment string and limited file fingerprints.
  14. Repository mutations are covered more broadly, but runtime/DB/evidence mutations can remain green.
  15. Set equality passes, but the denominator itself is not independently established and assertions are not fully audited.
  16. Yes, based on regression checks. T0F-002 and T0F-013 remain closed.
  17. No. The manifest is demonstrably not bound to the exact Golden runtime state.