KB-2DC8

06 — No-Hardcode / No-Local-Governance-Island Check + Forbidden-Compliance (Round 4 Concept Patch, 2026-06-01)

6 min read Revision 1
one-roof-governanceround4concept-patchno-hardcodeno-local-islandforbidden-complianceno-mutationno-version-bumpfive-layer-sync2026-06-01

06 — No-Hardcode / No-Local-Governance-Island Check + Forbidden-Compliance

Verifies this patch against the constitutional guardrails (prompt-muc-tieu-mo §5: no_hardcode_absolute, discover_first_reuse_first, five-layer sync, Design-Only Macro Mode) and the mission §4 forbidden list.

6.1 No-hardcode check (the patch content itself)

The concept layer is the anti-hardcode layer; the patch was checked so it does not itself introduce a hidden hardcode.

Check Result
No fixed axis array introduced ✅ — open-axis model (M-DEF-8/9) is explicitly "no fixed axis array anywhere"; future axis = Axis-Registry row, not code.
The one existing substrate hardcode is surfaced, not hidden ✅ — SB-3 (iu_three_axis_envelope hardcodes 3 axes) is named in canon 01 §8, 02 §3, 03, Reg-Pivot 07, and the IU patch. The patch states "no fixed axis array" concept-true, substrate-false until SB-3 generalized — it does not pretend the substrate is already open.
No fixed object-class list pretending to be code ✅ — 13 classes + Class 0 are a data registry (M-DEF-2); new type = data row.
No literal owner/threshold baked into a concept ✅ — ownership is relational (no owner_gov_code column; canon 01 §2); the 50-ceiling is a per-species policy value in the Axis Registry, not a hardcoded constant (Reg-Pivot 07 patch).
count>1 not hardcoded as a mandate ✅ — M-DEF-10 makes it a candidacy trigger; the rule is data/predicate-driven (shared-truth + grain), not a fixed action.

6.2 No-local-governance-island check

Check Result
Did the patch create a second governance roof? ✅ No — the concept canon centralizes under Điều 37 (one roof); surfaces reference it (anti-bloat), they do not mint local owners/approvals.
Did any patched surface doc declare a local owner/approval/lifecycle? ✅ No — Reg-Pivot 00 explicitly states the surface "mints no local owner/approval/policy/lifecycle" and that any such declaration = LOCAL_GOVERNANCE_ISLAND.
Is the island definition itself carried (dual PG + CI detection)? ✅ — canon 01 §5; Reg-Pivot 06 patch names both channels.
Is the new concept canon itself an island? ✅ No — it is the Điều 37 OWN layer (the central definitions), not a parallel governance authority. It binds no owners and creates no approvals; it is a concept proposal awaiting the council/human surface.
IU island risk ✅ Carried, not created — the IU patch states IU "must not maintain a governance island" and defers owner-binding (OP-B) rather than minting a local IU owner. The live 54-command dot_iu_command_catalog island and free-text owner_ref are flagged as the existing condition OP-B must resolve, not endorsed.

6.3 discover_first / reuse_first

✅ The patch discovered the canonical design home (knowledge/dev/design/) and the real targets before editing (doc 02). It reuses the existing surface docs (append-only concept sections) and the existing Round-4 clause cards (transcribed into the canon), rather than authoring duplicate governance content. Ownership stays relational (reuses governance_registry + relations + law_jurisdiction), not new columns.

6.4 Five-layer sync impact

None. The patch touches only the AgentData/KB layer (concept design docs). PG, Directus, Nuxt, and Qdrant are untouched. There is no cross-layer divergence to reconcile because no other layer changed.

6.5 Forbidden-action compliance (mission §4 + §6)

Forbidden Complied?
PG schema implementation ✅ none
DOT implementation ✅ none
scanner SQL / view / function ✅ none
issue / event registration ✅ none
apply / remediation binding ✅ none
object/axis ownership table design beyond approved concept references ✅ none — table is referenced as a C-1/SB-2 blocker, not designed
IU surface / binding technical design ✅ none — IU concept-only; OP-B/SB-3 held
production UI / API / route ✅ none
law enactment ✅ none
normative_registry / law_catalog / governance_docs write ✅ none
approval / self-approval ✅ none
version numbers / status change ✅ none (see §6.6)
anything that pretends blockers are solved ✅ none — §3.7 deferred-technical declaration + every Blockers line
PG / Directus / Qdrant / Nuxt mutation ✅ none — KB document patch/create only

6.6 The "no version bump" nuance (transparency)

patch_document advanced each edited doc's KB revision field (1→2). This is the KB store's automatic per-edit counter — it is not a semantic version bump:

  • No design document's declared version was changed (Registries-Pivot stays v0.1; IU foundation carries no version; the canon docs are new, no prior version).
  • No status line was changed (all remain DESIGN-ONLY / DRAFT / NOT ratified).
  • No law version (Đ37 v3.3, etc.) was touched; normative_registry was not written.

The revision increment is unavoidable when appending a section and is recorded here for completeness, not as a violation of "no version bump."

6.7 No-mutation confirmation

  • PG / Directus / Qdrant / Nuxt: zero mutation — no such tool invoked for writes this session.
  • KB: 13 writes = 4 concept-canon upload_document (new) + 9 report upload_document (new) + 7 patch_document (append-only sections on existing canonical docs). No delete; no existing content removed.
  • Net effect on system truth = zero new enacted/ratified state. Additive concept-design + reporting only.
Back to Knowledge Hub knowledge/dev/reports/architecture/patch-canonical-design-one-roof-concepts-round4-2026-06-01/06-no-hardcode-no-local-governance-check.md