12 — Next Paste-Ready Prompts

Eight self-contained prompts. Each begins with state recovery + Hard Gate 0 and carries its own forbidden list, per the mission standard. Default next step = Prompt 1.

---

Prompt 1 — CONDITIONAL-GO scoped canonical-design concept patch (the GO path)

MISSION: PATCH_CANONICAL_DESIGN_ONE_ROOF_CONCEPTS_ROUND4
Effort xhigh. Read-only KB + live PG read; apply channel = canonical DESIGN docs only (concept tier). NO PG/Directus/Qdrant/Nuxt mutation, NO law file edit, NO enactment, NO version bump, NO approval.
Recover state from knowledge/dev/reports/architecture/one-roof-governance-law-hardening-finalization-round4-2026-06-01/ (docs 00–13).
Patch ONLY the 12 concept/principle canonical design docs listed in doc 11 §11.2, each carrying the clause cards from doc 10 (M-DEF-1..10 + §0-GOV + the specialized-law reference pattern) and EACH explicitly flagging its blocker(s): SB-1/SB-2/SB-3, OP-B, L-1/L-2.
HARD CONSTRAINTS: do NOT patch apply/remediation-binding docs; do NOT patch IU owner-binding/surface docs (OP-B held); do NOT patch any law file; do NOT generalize the IU axis envelope; carry every "no fixed axis array" claim WITH the SB-3 caveat.
Deliver: the patched concept docs + a diff manifest + an acceptance check that each doc carries its M-DEF references, its §0-GOV block, and its blocker flag. Self-review + GO/NO-GO for the next (apply) tier.
Forbidden: enactment, version bump, approval, PG/UI/route/schema mutation, hardcode, fixed axes, hidden island, redesign of settled defs.

Prompt 2 — Council/GPT adjudication of the 8-item decision surface

MISSION: COUNCIL_ADJUDICATE_ONE_ROOF_DECISION_SURFACE_ROUND4
Read-only. Recover doc 02 (open-question closure) from the Round-4 package.
For EACH of the 8 council/human items (C-1 object-edge table, C-2 APR action-type bundle, C-3 OP-B IU owner, C-4 review_decision adapter, C-5 GOV-MOUT activation, C-6 A2/A3 boundaries, H-1 enactment phase, H-2 OP-B ratification): present the recommended default, the risk, the alternatives, and request a RULE/DEFER decision. Produce a decision-record doc per item (do NOT enact; record decisions as draft minutes only).
Forbidden: enactment, approval creation, PG mutation, version bump, self-approval.

Prompt 3 — Fifth hardening round (contingency if review rejects CONDITIONAL GO)

MISSION: ONE_ROOF_GOVERNANCE_LAW_HARDENING_ROUND_5
Use ONLY if a reviewer downgrades Round 4 to NO-GO. Read-only. Recover the Round-4 package.
Take the reviewer's exact objections; for each, either fold a wording fix into a revised clause card (doc 10) or convert it to a named blocker with default. Re-run Red-Team (≥64) and re-issue GO/NO-GO. Do NOT restart; fold only.
Forbidden: same as Round 4 §18.

Prompt 4 — Content-only law cleanup + registration

MISSION: CONTENT_ONLY_LAW_CLEANUP_AND_REGISTER_ROUND4
Read-only survey + DRAFT-only law-text corrections (no enactment, no version bump, no normative_registry write — DRAFT packet only).
Resolve L-1 drift: Đ36 v4.0-vs-v5.0 authoritative version; Đ45 strip `ban_hanh=false` leftover; Đ37 v3.3 text reconciled to live columns (gov_type/created_by_law/relation_type/enforcement_role + capability JSON).
Resolve L-2 registration gap: prepare normative_registry DRAFT rows for Đ20, Đ23, Đ44(draft), Đ45 (status per KB headers). Output a ratification packet (exact rows + diffs) for the human enact phase. Do NOT write to normative_registry/law_catalog/governance_docs.
Forbidden: enactment, version bump, registry write, approval.

Prompt 5 — Substrate design for SB-1/SB-2/SB-3 (design + read-only rehearsal)

MISSION: DESIGN_GOVERNANCE_SUBSTRATE_SB1_SB2_SB3_ROUND4
Design-only macro (mission Design-Only Mode). Read-only PG + BEGIN..ROLLBACK rehearsal ONLY if necessary; prefer read-only inspection. NO committed DDL/DML.
SB-1: design the 4 apr_action_types (assign_governance_owner+handler, grant_governance_exception, delegate_authority, assign_axis_owner) + handler/council-review wiring.
SB-2: design the additive governance_object_ownership table (owner_gov_code, object_type, object_ref, scope, capability_ref, effective_from/to, ...) — preferred over widening governance_relations CHECK; prove no migration risk to the 8 live edges.
SB-3: design the IU Axis Registry + generic iu_axis_value store + the iu_three_axis_envelope demotion to a hot-cache projection; prove a 4th axis becomes data, not DDL.
Deliver: DDL drafts (uncommitted), rollback scripts, a read-only rehearsal log, and an apply-readiness gate. Do NOT commit anything.
Forbidden: committed DDL/DML, enactment, approval, production mutation.

Prompt 6 — OP-B IU owner assignment council packet

MISSION: DECIDE_OP_B_IU_OWNER_ROUND4
Read-only. Recover doc 05 (IU hardening). Produce a council decision packet for OP-B: confirm/adjust the recommended owner-per-scope (policy→COUNCIL, substrate/health→KG-SYS+SIV, exec→DOT, render→MOUT-interim-delegation, law→NRM-SYS) and rule on C-4 (review_decision adapter vs full Đ32). Include the live evidence (no GOV-IU agency; owner_ref free-text; conformance_status='open' all 219; 54-cmd island). Output draft minutes only; do NOT bind ownership into governance_registry, do NOT enact.
Forbidden: PG mutation, approval, enactment, version bump.

Prompt 7 — Governance-coverage scanner / DOT technical design (ONLY after concept patch accepted)

MISSION: DESIGN_GOVERNANCE_COVERAGE_SCANNER_DOTS_ROUND4
PRECONDITION: Prompt 1 concept patch accepted. Design-only + read-only rehearsal.
Design the 6-layer coverage detector + the scan/propose/apply/audit DOT family (reuse the Đ19 orphan scanner + Đ35 DOT pattern + GOV-SIV); add dot_coverage_required rows (governance.coverage/classification/pivot/axis/iu) for SB-8; design the anti-spam path (coalesce_key/cooldown/summary/emit-ceiling) proven against the live template_gap=182,378 scale; design the inventory-completeness reconciliation (the governance twin of birth-orphan). Output design + read-only rehearsal; NO committed DOT, NO production scan emit.
Forbidden: committed DOT/scan, event emit, PG mutation, enactment.

Prompt 8 — Cowork / Claude Desktop / GPT critique

MISSION: COUNCIL_CRITIQUE_ONE_ROOF_ROUND4
External-reviewer critique of the Round-4 package. Attack: (a) does any closed open-question default leak a local island? (b) does the count>1 rule (M-DEF-10) over- or under-govern any real case? (c) is SB-3 (axis envelope) the only hidden hardcode, or are there others? (d) does the CONDITIONAL GO concept scope smuggle in any apply/owner-binding? (e) does Red-Team v3 miss an attack class? Return a defect list with severity + required fold; recommend GO / CONDITIONAL GO / NO-GO.

---

Sequence: Prompt 1 (concept patch) ∥ Prompt 4 (law cleanup) ∥ Prompt 6 (OP-B packet) ∥ Prompt 8 (critique) → Prompt 2 (council adjudication) → Prompt 5 (substrate design) → Prompt 7 (scanner design, gated) → [human enact phase] → Prompt 3 only if downgraded.