11 — T1 Implementation Boundary Review (SUPERTRACK K)

Source: 10-t1-implementation-boundary.md.

#	Requirement	Spec answer	Verdict
K.1	what T1 can author/test in FIX7a	repository files: exact DDL/functions/tests/manifests/packets/analyzer + local/rehearsal tests	PASS
K.2	what T1 cannot do	no production mutation, no activate/seal, no production-owner trigger, no permit/apply, no source/UI change, no choosing alternatives / no improvising missing spec	PASS
K.3	what requires operator	all live role/schema/extension/FIX7a/owner/ACL/activation/scheduler/repoint/FIX7b/c/rollback	PASS
K.4	what requires Codex re-audit	Codex re-audit before live FIX7b/c and after deploy	PASS
K.5	what remains blocked until 2.6B	Stage 2.6B/apply blocked	PASS
K.6	expected final status if owner/ACL not applied	gate stays FIX7_DESIGN_SPEC_READY_FOR_T1_REVIEW_NOT_IMPLEMENTATION; live cutover operator-gated; readiness stays false	PASS
K.7	exact self-audit dashboard required for T1	GAP — doc 10 does not specify the T1 self-audit dashboard/queries; prior FIX6 discipline (self-audit + independent adversarial sub-check) is implied but not made a required artifact	PARTIAL

Adversarial probes

• Is the boundary watertight against improvisation? Largely yes — FIX7_IMPLEMENTATION_BLOCKED_SPEC_CONFLICT requires T1 to STOP on any missing spec rather than improvise. This is the key safety valve. But its effectiveness depends on the byte-level spec being complete enough that T1 recognizes a gap as a conflict rather than plausibly filling it. With the manifest/child DDL, gate adapters, and hash key-maps unpublished (docs 02/03/07), T1 would hit SPEC_CONFLICT repeatedly — or, worse, author plausible-but-divergent artifacts without realizing it is a conflict (the exact FIX..FIX6 failure mode). This is why the package needs the artifact correction before T1 starts.
• Self-audit requirement: the macro (K.7) asks for "exact self-audit dashboard required for T1." Doc 10 does not define it. Recommend Codex add the required T1 self-audit dashboard (the FIX6 pattern: self-audit after build + independent adversarial read-only sub-agent; self_audit_pass + independent CONFIRM) as a named deliverable.

Verdict: T1_BOUNDARY_COMPLETE (with one gap: K.7 self-audit dashboard)

The author-vs-operator boundary and the SPEC_CONFLICT-stop discipline are clear and correct. Two refinements: (1) the boundary only holds if the byte-level spec is complete (cross-ref B/G); (2) add the explicit T1 self-audit dashboard requirement (K.7).