KB-5194

RS5B-CLOSEOUT 07 — P2 Entry Bad Inputs & Fail-Closed Oracles — 2026-06-21

10 min read Revision 1
rs5b-closeoutbad-inputsfail-closedxbiadversarialregistration-hold2026-06-21

RS5B-CLOSEOUT 07 — P2 Entry Bad Inputs & Fail-Closed Oracles — 2026-06-21

Purpose: adversarial bad inputs that test the closeout consolidation (Job A) and the P2 entry gate (Job B). Each fails closed. No invalid input produces a PASS, seal, digest, certificate, authority token, or registration-ready label. These are design fixtures (DEFINED_NOT_EXECUTED, per caveat C3), not executed tests. REGISTRATION_HOLD retained · 0 mutations.

Each fixture: input shape · expected rejection · why no other code competes · PASS/seal/digest possible? · result.


XBI-1 — RS5B accepted as runtime authorization

  • Input shape: a summary asserts "RS5B (as corrected) is accepted, therefore the founding act is authorized to execute / ownership may be minted."
  • Expected rejection: AUTHORITY_OVERCLAIM.
  • Why no other code competes: the defect is conflating contract-layer acceptance with runtime/execution authority. It is not a scope-drift (no registration act is attempted yet), not a LEGO failure, not a mutation. Only AUTHORITY_OVERCLAIM names "acceptance-of-design treated as authorization-to-execute" (caveat C5; file 02 §2 design-solved ≠ runtime-resolved).
  • PASS/seal/digest possible? No.
  • Result: REJECTED. Contract acceptance is necessary-not-sufficient; execution needs packet item 13 (Chairman). HOLD retained.

XBI-2 — P2 started from PATCH2 accept without consolidated closeout

  • Input shape: "Codex ACCEPT_RS5B_PATCH2 exists, so open the P2 build lane now."
  • Expected rejection: P2_OPENED_EARLY.
  • Why no other code competes: the act is opening P2, gated by R1 (accepted consolidated contract) + R12 (independent review of the entry design) + R2 (Chairman packet). It is not a generic overclaim of authority over an existing act, and no register_dot is attempted, so it is P2_OPENED_EARLY specifically — premature lane opening, not registration drift.
  • PASS/seal/digest possible? No.
  • Result: REJECTED. PATCH2 acceptance is contract-only (C5); P2 opens only after this closeout's entry-gate design is independently accepted plus Chairman authorization.

XBI-3 — canonical_operation vocabulary treated as a runtime-present row

  • Input shape: an act resolves canonical_operation against an assumed existing vocabulary table / inserts a vocabulary row to proceed.
  • Expected rejection: VOCABULARY_RUNTIME_OVERCLAIM (with underlying CANONICAL_OPERATION_VOCABULARY_REQUIRED_NOT_PRESENT).
  • Why no other code competes: the specific fault is treating a REQUIRED_NOT_PRESENT design surface as a runtime fact (or creating it without separate authorization). Distinct from WRONG_CANONICAL_OPERATION_FOR_EFFECT (which is using the wrong operation value, e.g. register_dot for a founding act) — here the vocabulary itself is asserted/created.
  • PASS/seal/digest possible? No.
  • Result: REJECTED. Vocabulary stays design-only (R5); no default to register_dot, no invented row, HOLD.

XBI-4 — BINDING_CHECK_PASS treated as registration PASS

  • Input shape: the BI/effect-binding classifier returns BINDING_CHECK_PASS, and the act proceeds to register/activate as if authorized.
  • Expected rejection: PASS_LEVEL_OVERCLAIM.
  • Why no other code competes: the binding check itself did pass (so it is not a binding reject like AUTHORIZATION_BINDING_MISSING_EFFECT); the defect is promoting a necessary-not-sufficient PASS to an authority/registration PASS. Only PASS_LEVEL_OVERCLAIM names that promotion.
  • PASS/seal/digest possible? No (the upstream BINDING_CHECK_PASS is not a seal; it is explicitly necessary-not-sufficient).
  • Result: REJECTED. BINDING_CHECK_PASS ↛ registration/authority/activation PASS; HOLD + all carried blockers remain.

XBI-5 — approval authority inherited into registration authority

  • Input shape: GOV-COUNCIL approval / DOT_APPROVAL_QUORUM_AUTHORITY is used as the authority to create the ownership/registration row.
  • Expected rejection: IMPLICIT_AUTHORITY_INHERITANCE_REJECTED (edge DOT_APPROVAL_QUORUM_AUTHORITY ↛ DOT_REGISTRATION_AUTHORITY).
  • Why no other code competes: the fault is an authority inheritance across scopes (R8 / Model C rejection), not premature lane opening (XBI-2) or generic overclaim (XBI-1). The named must-not-inherit edge makes the code unambiguous.
  • PASS/seal/digest possible? No.
  • Result: REJECTED. Approval-authority ≠ registration-authority ≠ activation-authority; approval may only check, never originate ownership.

XBI-6 — carrier bundle created as one mega-registry

  • Input shape: C1–C7 merged into a single table/graph/birth pipeline "for convenience," with shared lifecycle and no per-carrier rollback.
  • Expected rejection: LEGO_BOUNDARY_INSUFFICIENT.
  • Why no other code competes: the fault is structural boundary collapse (R7), independent of authority or registration. Even with perfect authority, a mega-registry violates born/test/change/rollback separation.
  • PASS/seal/digest possible? No.
  • Result: REJECTED. Each carrier must be born/tested/changed/rolled-back separately, joined only by explicit reference edges (file 06).

XBI-7 — rollback omitted for any carrier

  • Input shape: a carrier (any of C1–C7) is proposed for build without a per-carrier rollback plan.
  • Expected rejection: ROLLBACK_PLAN_ABSENT.
  • Why no other code competes: the missing element is specifically the rollback surface (R6), not the LEGO partition as a whole (XBI-6) — a carrier can be properly partitioned yet still lack a rollback plan.
  • PASS/seal/digest possible? No.
  • Result: REJECTED. No carrier without a rollback surface may enter P2 build.

XBI-8 — unsafe registrar wrapper / reuse proposed

  • Input shape: P2 proposes to wrap/relabel/reuse the dot-dot-register mass-scan real-run path instead of the DOT_REGISTER_GOVERNED_REPLACEMENT identity.
  • Expected rejection: UNSAFE_REGISTRAR_REUSE_REJECTED.
  • Why no other code competes: the fault is reuse of a known-unsafe execution path (R9 / replace-not-wrap), not registration admission itself (XBI-9) — the wrapper could exist before any admission attempt.
  • PASS/seal/digest possible? No.
  • Result: REJECTED. Replace-not-wrap; allowed reuse limited to dry-run / classify-advisory / source / warning surfaces.

XBI-9 — register_dot admission attempted in P2

  • Input shape: P2 performs a real register_dot admission of a DOT.
  • Expected rejection: P2_SCOPE_DRIFT_REGISTRATION_ATTEMPT.
  • Why no other code competes: the act is a real registration inside the build-design lane (R3/R10). It is not mere lane-opening (XBI-2) and not registrar reuse (XBI-8) — it is the registration act itself, which is a P3-only act behind a separate later gate.
  • PASS/seal/digest possible? No.
  • Result: REJECTED. register_dot real admission stays in P3 behind its own gate, after all pre-runtime carriers exist + PASS.

XBI-10 — runtime write or DDL/DML appears

  • Input shape: any DDL/DML, PG/Directus mutation, row insert/update/delete, or handler write surfaces anywhere in closeout or P2-entry processing.
  • Expected rejection: RUNTIME_MUTATION_REJECTED.
  • Why no other code competes: this is the top-priority short-circuit — any observed runtime mutation is rejected regardless of which other predicate might also match. It dominates all XBI codes (the gate-order short-circuit in file 05 §2).
  • PASS/seal/digest possible? No.
  • Result: REJECTED. This package and the P2 entry gate are KB-only / design-only; runtime mutation is categorically forbidden until a separately-authorized lane under Chairman authority.

2. Oracle totality (classifier-scoped)

The XBI oracle maps each declared bad input to exactly one code from the closed set:

{ AUTHORITY_OVERCLAIM, P2_OPENED_EARLY, VOCABULARY_RUNTIME_OVERCLAIM, PASS_LEVEL_OVERCLAIM,
  IMPLICIT_AUTHORITY_INHERITANCE_REJECTED, LEGO_BOUNDARY_INSUFFICIENT, ROLLBACK_PLAN_ABSENT,
  UNSAFE_REGISTRAR_REUSE_REJECTED, P2_SCOPE_DRIFT_REGISTRATION_ATTEMPT, RUNTIME_MUTATION_REJECTED }

Per caveat C2 this totality is scoped to this declared classifier and its inputs, not to arbitrary malformed serialization or implementation behavior. Precedence: RUNTIME_MUTATION_REJECTED short-circuits first; otherwise the lowest-numbered failing entry-gate requirement (file 05 §2) determines the code. Consistent with the closeout's allowed verdict set: XBI-1↔RS5B_CLOSEOUT_REJECT_AUTHORITY_OVERCLAIM; XBI-2↔RS5B_CLOSEOUT_REJECT_P2_OPENED_EARLY; XBI-6↔RS5B_CLOSEOUT_HOLD_LEGO_BOUNDARY_INSUFFICIENT; XBI-9/XBI-3 scope drift↔RS5B_CLOSEOUT_HOLD_P2_ENTRY_SCOPE_TOO_BROAD; XBI-10↔RS5B_CLOSEOUT_REJECT_RUNTIME_MUTATION.

3. Self-result

All ten XBI fixtures fail closed. None yields a PASS/seal/digest/certificate/authority-token/registration-ready output. The closeout itself does not match any XBI reject (no runtime mutation, no authority overclaim, no early P2 opening, LEGO boundaries shown per carrier). Status: CLOSEOUT_ADVERSARIAL_SELF_CHECK_PASSED_NO_FAIL_OPEN.

Back to Knowledge Hub knowledge/dev/laws-new/reports/rs5b-closeout-p2-entry/07-p2-entry-bad-inputs-and-fail-closed-oracles-2026-06-21.md