docs(r4): full-stack omni-emission canvas (TS + React substrate)

[briansrls]

Summary

• R4 canvas (post-R3 scope) per Director ratification msg_22a1c596 relayed via PM msg_83ce8113 (2026-05-13) + operator directive
• Surfaces 5 Director-framed sub-canvas questions Q1-Q5 + Practice 4 disposition on HookKind
• Substrate audit confirms TS is genuinely net-new; gate Fix Value serialization and implement is_empty() method #28 omni_layers_share_one_node_tree provides cross-target invariant extension point
• Hard-bound: canvas-only; NO implementation pre-R3 close
• Companion: Director-owned path (a) 4-layer TODO demo using existing substrate

Mgr recommendations (summary)

Q	Disposition	Rationale
Q1 TS shape	Q1-b (structural-vs-nominal axis on InhabitantDecl)	encodes genuine semantic difference (Practice 2/6)
Q2 React shape	Q2-a (Shape-A)	components ARE TS source code; no new shape category needed
Q3 ingest	Q3-a (.dag → JSX)	single-authority discipline
Q4 #28 invariant	Director disposition required (extend vs new gate)
Q5 lens compose	Q5-a (Component as Behavior::Bind)	strict-mirror; lens framework applies uniformly

Test plan

• docs-only change; relies on CI fmt/changes
• Director ratification on §12 Q1-Q5 + Practice 4 disposition on HookKind

🤖 Generated with Claude Code

[briansrls]

cursor/composer-2 review 10817 — APPROVE w/ exploratory observation addressed in commit 5b7c00e.

§8 Practice 4 table previously framed HookKind as "7-arm closed enumeration" while §12 separately framed "6 standard hooks + Custom(Identifier)". Reframe §8 to match §12: 6 standard-hook arms (UseState/UseEffect/UseMemo/UseCallback/UseContext/UseRef) covering React 18 + 1 user-input boundary arm (Custom(Identifier)); Custom requires Practice-4-promotion canvas per §10 Phase-1.5.

Eliminates two-different-coproduct-sizes reading.

— sent from warm-wolf-698

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: d251b61c · Trigger: schedule
• Thinking: 223s wall

---

BLOCKING (2)

Root Cause

• docs/design-r4-full-stack-omni-emission-canvas.md React hook substrate is authored from an incomplete hook roster → cite the React versioned hook authority and either enumerate all built-ins for that version or narrow the claim to a bounded subset with a dissolution trigger.
• docs/design-r4-full-stack-omni-emission-canvas.md The carrier sketch treats subcomponent use as an alternate body mode instead of a node/edge inside JSXTree → model Component.body as one render tree with component references as tree nodes or explicit derived coordinates.

⚠️ The canvas needs those substrate-shape corrections before it becomes the R4 worker authority.

[briansrls]

codex BLOCKING review d251b61 — both findings addressed in commit 945fa96:

Finding 1 (HookKind incomplete roster): Previous 6-arm enumeration under-counted React 18 built-in hooks. Fix-forward: 15-arm closed enumeration of all React 18.3 built-ins (authority anchor: react.dev/reference/react) — UseState/UseReducer/UseEffect/UseLayoutEffect/UseInsertionEffect/UseContext/UseRef/UseImperativeHandle/UseMemo/UseCallback/UseDebugValue/UseDeferredValue/UseTransition/UseId/UseSyncExternalStore + Custom(Identifier) boundary arm. Dissolution trigger: React version-anchor change (new 18.x/19.x built-in) re-ratifies roster on the version-anchor PR.

Finding 2 (ComponentBody subcomponent-as-alternate-mode): Previous ComponentBody = Render | Composite treated subcomponents as alternate body. Fix-forward: Component.body IS a JSXTree; subcomponents are JSXNode.ComponentRef nodes within the tree:

JSXNode = HtmlElement | ComponentRef | TextNode | ExpressionSlot | FragmentNode

Dissolves prior Render/Composite split — one render tree with component references as tree nodes.

§8 Practice 4 table + §12 ratification narrative updated. Surfacing to PM in parallel — these are material carrier-sketch corrections post-Director-ratification; may warrant brief re-acknowledgment that the §3 carrier sketch revision composes with the Q1-b/Q2-a/Q5-a dispositions (which it does — JSXTree-as-body is more strictly Q2-a Shape-A; 15-hook roster is more concretely Q5-a Behavior::Bind-applicable).

— sent from warm-wolf-698

[briansrls]

Re inline BLOCKING at canvas:168 (HookKind faithfulness) — already addressed in commit 945fa96 (pushed before this relay landed). HookKind expanded from 6-arm to 15-arm React 18.3 built-in roster with authority anchor react.dev/reference/react:

type HookKind =
    UseState | UseReducer | UseEffect | UseLayoutEffect | UseInsertionEffect
  | UseContext | UseRef | UseImperativeHandle | UseMemo | UseCallback
  | UseDebugValue | UseDeferredValue | UseTransition | UseId | UseSyncExternalStore
  | Custom(Identifier)

Dissolution trigger: React version-anchor change (new built-in in 18.x or 19.x) → re-ratify roster on the version-anchor PR. P1 / extdeps-fidelity restored; no first-party hook falls into Custom on React 18.3.

— sent from warm-wolf-698

[briansrls]

Re inline BLOCKING at canvas:169 (ComponentBody non-terminal coproduct) — already addressed in commit 945fa96 (pushed before this relay landed). The Render | Composite split is dissolved:

type Component { name, props, state, body: JSXTree }
type JSXTree { root: JSXNode }
type JSXNode =
    HtmlElement { tag, attrs, children }
  | ComponentRef { component, props, children }
  | TextNode { text }
  | ExpressionSlot { expr }
  | FragmentNode { children }

Subcomponents are now JSXNode.ComponentRef tree-nodes inside Component.body — mixed render/composition structure is representable as a single JSX tree containing both HtmlElement and ComponentRef arms at any depth. Practice 4 satisfied; no parallel-authority split.

— sent from warm-wolf-698

[briansrls]

cursor 10851 APPROVE_WITH_COMMENTS — both findings fixed in commit 6590d74:

1. §6 L226 Shape-F stray: "React UI (new Shape-A or Shape-F)" → "new Shape-A per ratified Q2-a; Shape-F explicitly REJECTED (see anti-pattern §11 feat(cloud): add cloud resource management layer with GCP and AWS sup… #6)". Single-authority restored — no downstream brief can cite Shape-F as a valid path.

2. §3 L124 self-referential typo: ComponentRef { component: ComponentRef, ... } was a recursive name collision. Renamed the JSXNode arm to ComponentRefNode with field component: ComponentName — a distinct handle type referencing the named Component, not the JSXNode arm itself. Cascaded through §3 comment + §8 Practice 4 table.

— sent from warm-wolf-698

[briansrls]

codex REQUEST_CHANGES review 10864 — TypingDiscipline implicit-default fallback FIXED in commit 68cbd16.

The §12 Q1 disposition text "Rust/Python/Go default to Nominal" was the exact P3/Practice-6 violation flagged: missing field interpreted as plausible-value reintroduces convention-fallback semantics instead of API-level enforcement. The Q1-b ratification stands; only the migration shape tightens to fail closed.

Migration is now atomic + fail-closed across §3 / §10 / §11 / §12:

• Carrier extension PR MUST set typing_discipline = Nominal explicitly on every existing inhabitant (Rust + Python + Go) in the same atomic PR
• Carrier extension PR MUST add a compile-time exhaustiveness test asserting no InhabitantDecl row lacks the field
• Missing field MUST fail compilation — no implicit fallback to a plausible value
• New TS rows set Structural explicitly

Added §11 #10 (Mgr-derived anti-pattern): implicit-default landing for existing rows is forbidden.

— sent from warm-wolf-698

[briansrls]

cursor APPROVE 10884 — both exploratory observations addressed in commit 275bc71:

1. L83 Q1-b Cons "lazy migration acceptable": contradicted §12 ratified atomic+fail-closed disposition. Reworded to match — "ratified migration is atomic + fail-closed, NOT lazy" with cross-reference to §12 + anti-pattern §11 Design: Diff Gist Mode for Branch Change Snapshots #10. Skim-then-miss-override risk eliminated.

2. L313 Q3-a INVARIANTS P1 → P2 citation: "exactly one authoritative place" is P2 Boundary Discipline (INVARIANTS.md:144), not P1 Modeling Faithfulness. Fixed.

— sent from warm-wolf-698

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 39ffb8e6 · Trigger: schedule
• Thinking: 370s wall

---

BLOCKING (2)

Root Cause

• docs/design-r4-full-stack-omni-emission-canvas.md React hook/effect API signatures are flattened into generic Hook/Lifecycle records → derive lifecycle/effect facts from versioned hook invocation signatures and place dependency/cleanup fields only on hooks that accept them.

⚠️ Design/docs PR needs the React hook/effect carrier shape corrected before it becomes the R4 worker authority.

[briansrls]

BLOCKING: Hook.dependencies is attached to every HookKind, so non-dependency hooks like UseState can carry dependency facts while effect/memo/callback hooks lose their distinct call signatures, violating P1/P2/P6.

[briansrls]

BLOCKING: Lifecycle is classified GREEN, but OnMount/OnUnmount/OnUpdate are derived from React effect dependency and cleanup structure rather than irreducible lifecycle variants, so this should dissolve into Effect/Hook facts under Practice 4/P1.

[briansrls]

Re inline BLOCKING at canvas:182 (HookKind 6-arm React coverage incomplete) — already addressed in commit 945fa96 (pushed 2026-05-13 before this batch relay).

Current HEAD canvas:111-115 enumerates the full React 18.3 built-in roster:

type HookKind =
    UseState | UseReducer | UseEffect | UseLayoutEffect | UseInsertionEffect
  | UseContext | UseRef | UseImperativeHandle | UseMemo | UseCallback
  | UseDebugValue | UseDeferredValue | UseTransition | UseId | UseSyncExternalStore
  | Custom(Identifier)

15-arm closed enumeration covering all React 18.3 built-in hooks (authority anchor: react.dev/reference/react) + Custom(Identifier) user-input boundary. Dissolution trigger: React version-anchor change (new built-in in 18.x or 19.x) re-ratifies the roster on the version-anchor PR. P1 / extdeps-fidelity restored; no first-party React 18.3 hook falls into Custom.

Identical reply to the prior relay at the same finding (comment 4437821229). The batch reviewer is reading a pre-945fa96758 SHA. Will verify the remaining +4 queued against current state as they relay.

— sent from warm-wolf-698

[briansrls]

Re inline BLOCKING at canvas:183 (ComponentBody non-terminal coproduct) — already addressed in commit 945fa96 (with the ComponentRefNode rename in 6590d74).

Current HEAD canvas:106 + :121-127:

type Component { name, props, state, body: JSXTree }
type JSXTree { root: JSXNode }
type JSXNode =
    HtmlElement { tag, attrs, children }
  | ComponentRefNode { component: ComponentName, props, children }
  | TextNode { text }
  | ExpressionSlot { expr }
  | FragmentNode { children }

Subcomponents are JSXNode.ComponentRefNode tree-nodes inside Component.body — mixed render/composition structure is representable as a single JSX tree containing both HtmlElement and ComponentRefNode arms at any depth. Practice 4 satisfied; no parallel-authority split.

Stale-SHA batch relay; identical to comment 4437823062.

— sent from warm-wolf-698

[briansrls]

Re inline BLOCKING at canvas:76 (Hook.dependencies on uniform record) — FIXED in commit bde92bd. Good catch.

The reviewer is right: uniform Hook.dependencies let UseState/UseRef/UseContext (which don't take dependency arrays) carry meaningless dependency facts AND erased distinct call signatures of effect/memo/callback/imperative-handle hooks. P1/P2/P6 violation.

Fix-forward: drop Hook.dependencies; move call-signature fields into each HookKind arm per React 18.3 reference. Each arm now carries exactly the fields its hook takes — UseState{initial}, UseEffect{body, dependencies, cleanup?}, UseReducer{reducer, initial}, UseImperativeHandle{ref, factory, dependencies}, UseSyncExternalStore{subscribe, get_snapshot, get_server_snapshot?}, etc. UseTransition / UseId carry no fields. Hook record reduces to { name, kind: HookKind }. Standalone Effect type dropped (cleanup now on UseEffect arm directly).

Added §11 #11 anti-pattern: call-signature fields on uniform Hook record are forbidden — they belong on the per-arm carrier.

— sent from warm-wolf-698

[briansrls]

Re inline BLOCKING at canvas:184 (Lifecycle non-irreducible coproduct) — FIXED in commit 802a322.

Reviewer correct: OnMount/OnUnmount/OnUpdate are NOT irreducible — they derive from UseEffect arm structure:

• OnMount ≡ UseEffect { dependencies: [], cleanup: None }
• OnUnmount ≡ UseEffect { cleanup-only }
• OnUpdate(triggers) ≡ UseEffect { dependencies: triggers }

Parallel-authority sum alongside UseEffect violates Practice 4 / P1. Fix-forward:

• §3 sketch: Lifecycle DROPPED with explicit dissolution receipt comment showing each variant → UseEffect arm shape
• §8 Practice 4 table: Lifecycle reclassified 🔴 RED → DISSOLVED with cite to this finding
• §2 substrate audit: clarified Lifecycle + Effect NOT introduced (both dissolve)
• §11 Refactor tool acquisition: env node provides resources via edges #12 (new Mgr-derived anti-pattern): forbid parallel Lifecycle sum; lifecycle reasoning is a derived projection of UseEffect facts

— sent from warm-wolf-698

[briansrls]

Re BLOCKING codex 39ffb8e (Hook/Lifecycle flattened API signatures) — already addressed in commits bde92bd + 802a322 (pushed before this batch relay).

Both substrate-shape corrections are landed:

bde92bd — per-arm HookKind call signatures: dropped uniform Hook.dependencies; each React 18.3 built-in arm carries exactly its call signature:

• UseState { initial }, UseReducer { reducer, initial }
• UseEffect / UseLayoutEffect / UseInsertionEffect { body, dependencies, cleanup? }
• UseContext { context_ref }, UseRef { initial }
• UseImperativeHandle { ref, factory, dependencies }
• UseMemo { factory, dependencies }, UseCallback { callback, dependencies }
• UseDebugValue { value, format? }, UseDeferredValue { value }
• UseTransition (no args), UseId (no args)
• UseSyncExternalStore { subscribe, get_snapshot, get_server_snapshot? }
• Custom(Identifier)

Hook reduces to { name, kind: HookKind }. Dependency fields ONLY on hooks that accept them.

802a322 — Lifecycle dissolved into UseEffect arm structure: OnMount/OnUnmount/OnUpdate are derived projections of UseEffect, not irreducible variants. Lifecycle sum dropped; dissolution receipt added.

Anti-patterns §11 #11 + #12 added to forbid future regression to flattened shape.

— sent from warm-wolf-698