docs(briefs): R3 Cluster M Phase 2/3 dispatch briefs (#87 cementing-worker + #84 bulkport-coordinator)

[briansrls]

Closes #1940

Summary

Two new dispatch briefs for R3 Cluster M Phase 2 + Phase 3, bundled per feedback_bundle_workstreams_per_pr ratification at gunbc#2075 c#4413268466:

• Phase 2 worker brief (docs/briefs/r3-v-cluster-m-87-cementing-worker.md, +99): light port of multi-gate PRE-AUTH r3-v-tests-as-data-v1-worker.md narrowed to gate RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 (lens_cementing_test_discipline_complete). Discipline pattern (DifferentialEquals for v2-counterpart lenses, LensOutputEquals for v3-native), first-migration target, dispatch-ratchet successor, receipt + ledger updates. Independent of Cluster M Phase 1 per codex BLOCKING Consolidate binaries into gunbc-dag package #4 authority correction on PR docs(briefs): R3 Cluster M dispatch briefs (Task 2 per Director γ-hybrid ratification) #2362 (cementing axis orthogonal to property-based ProgramGenerator/Quantifier axis).

• Phase 3 coordinator skeleton (docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md, +88): 6-class brief queue (cementing / reflected-Dag / generic-DimReport / boundary / R1C-D-E / L4-L7-L5); strict-zero close-condition memorialized per Director Ask 4 (no closure-allowed exceptions; bulk-port = full EXPECTED_HAND_AUTHORED_TEST); lane-Mgr signoff workflow; per-class brief authoring discipline. Per-class detail authored as Phase 2 mid-flights.

Cite-and-execute discipline holds — substrate-of-truth lives at docs/design-tests-as-data-completeness.md §5 + §C5; briefs cite, do not restate.

Authority chain:

• Director ratification gunbc#846 #issuecomment-4412309986 (Asks 1/3/4: γ-hybrid / Verification coordinator / strict-zero).
• Cluster M dispatch overlays (Phase 2 RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 + Phase 3 Implement SDLC pipeline: worker dispatch, stage handlers, and integration tests #84) at PR docs(briefs): R3 Cluster M dispatch briefs (Task 2 per Director γ-hybrid ratification) #2362.
• Director substantive ratification of both briefs at gunbc#2075 c#4413268466.

Test plan

• Docs-only change — no code touched.
• Both briefs cite parent authority docs (cite-and-execute; no substrate-of-truth restatement).
• R1C-D/E pre-Phase-1 pilot pre-cleared per Director c#4413268466 (dispatched as follow-on per-class brief post-merge).
• No SG-0 hand-Rust authoring; SG-0 hand-path delta: +0.

🤖 Generated with Claude Code

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: f5c49b56 · Trigger: manual
• Comparison: main @ 7e043998 ... session/wise-bear-525 @ 8408f89b
• Conversation: View conversation

---

1. Story of the diff

This PR adds two docs-only dispatch briefs that stage Cluster M’s test-side dissolution path rather than changing compiler behavior. The #87 worker brief narrows the broader tests-as-data plan to a first cementing-test migration: use existing DifferentialEquals / LensOutputEquals predicate shapes, land one .dag TestClaim, delete the corresponding hand-Rust cementing test, then replace the current Rust cementing dispatch ratchet with a .dag claim-list ratchet. docs/briefs/r3-v-cluster-m-87-cementing-worker.md:19, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:22, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:63, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:65, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:72

The #84 coordinator brief then frames Phase 3 as the bulk-port follow-through: no closure-allowed exceptions, all live EXPECTED_HAND_AUTHORED_TEST entries must reach zero, and remaining test families are split into per-class worker batches with lane-manager behavioral signoff. It deliberately avoids hardcoding the live census count and points future workers back to locked design / live authorities instead of creating a second source of truth. docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:22, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:25, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:33, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:40, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:54, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:84

2. Invariant categories

1. LAYER MODEL — Compliant. This diff does not add substrate/Dag types or variants; the RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 brief explicitly forbids inventing new TestPredicate variants and directs workers to existing DifferentialEquals / LensOutputEquals, while Implement SDLC pipeline: worker dispatch, stage handlers, and integration tests #84 gates later bulk work on separately-landed Fix gist #85/BT1: SDLC pipeline compiles with unit_test profile #86 substrate carriers rather than smuggling that modeling into this brief. docs/briefs/r3-v-cluster-m-87-cementing-worker.md:22, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:44, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:45
2. INVARIANTS.md + modeling-discipline.md — Compliant. Boundary discipline / single authority is handled by keeping EXPECTED_HAND_AUTHORED_TEST as the live authority and refusing a hardcoded duplicate count; fail-closed/progress discipline is handled by requiring strict zero rather than closure exceptions, and by treating testgen gaps as Cluster M scope expansion rather than carve-outs. docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:22, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:24, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:25
3. CODING.md — N/A. The diff introduces no Rust functions, helpers, methods, result shapes, or implementation modules; the closest coding-style concern is dependency explicitness, and both briefs expose their authority lists up front. docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:7, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:7
4. TESTING.md — Compliant. No executable test is required for a docs-only dispatch brief, but the RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 acceptance criteria correctly preserve the cementing-test discipline: add a .dag TestClaim, delete the hand-Rust cementing file rather than stub-replacing it, decrement SG-0 by one, and require the runner to evaluate the claim green. docs/briefs/r3-v-cluster-m-87-cementing-worker.md:64, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:65, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:66, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:67
5. LOCKED DESIGN DECISIONS — Compliant. The briefs reference locked design surfaces and use cite-and-execute language rather than redefining them; RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 also says its claim shape mirrors the locked design sketch rather than replacing it. docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:10, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:84, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:8, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:13, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:44
6. TRACKED vs UNTRACKED DEBT — Compliant. The new skeleton/TBD surfaces are tracked: Implement SDLC pipeline: worker dispatch, stage handlers, and integration tests #84 labels itself as a coordinator skeleton, bounds the queue by class, names dispatch triggers, and gives close/ledger transitions; RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 similarly bounds the first-receipt slice and names the follow-on dispatch-ratchet successor before Phase 3 bulk-port begins. docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:3, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:27, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:44, docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:62, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:19, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:78, docs/briefs/r3-v-cluster-m-87-cementing-worker.md:83

3. Verdict

APPROVE

I found no reportable findings in the changed lines. The PR is docs-only, and the new briefs preserve single authority, strict-zero closure, and tracked-dissolution discipline rather than introducing new substrate or implementation behavior.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 8408f89b · Trigger: schedule
• Thinking: 238s wall

---

BLOCKING (2)

Root Cause

• docs/briefs/r3-v-cluster-m-87-cementing-worker.md The dispatch overlay's terminal reclassification was copied instead of preserving the locked substrate maturity annotations → cite these predicates as existing 🟡 scaffolds and keep their dissolution triggers visible.
• docs/briefs/r3-v-cluster-m-87-cementing-worker.md The v3-native row translated conceptual labels from the design prose instead of the live verification.dag field names → use LensOutputEquals { lens_ref, input_ref, expected_ref }.

Non-blocking — Strengths

• docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md The #84 coordinator brief preserves the Pure Bootstrap zero-floor by keeping EXPECTED_HAND_AUTHORED_TEST count = 0 as the strict PASSING gate.

⚠️ The docs need the predicate maturity and LensOutputEquals API corrected before dispatching workers from this brief.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 8408f89b · Trigger: manual
• Comparison: main @ 7e043998 ... session/wise-bear-525 @ 8408f89b
• Conversation: View conversation

---

1. Story of the diff

This PR adds two coordination briefs that turn the R3 Cluster M tests-as-data plan into dispatchable work. The #87 worker brief narrows the cementing-test discipline lane to a first concrete migration: use existing DifferentialEquals / LensOutputEquals predicates, add one .dag TestClaim, delete the corresponding hand-Rust cementing test, decrement SG-0, then replace the current Rust registry-dispatch ratchet with a .dag claim-list ratchet (docs/briefs/r3-v-cluster-m-87-cementing-worker.md:19, :63-68, :72-78). The #84 coordinator skeleton then frames the larger Phase 3 bulk-port: it keeps the SG-0 census as the live authority, splits the remaining hand-authored test population into per-class worker batches, and gates Phase 3 on #85/#86/#87 being landed (docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:22-25, :31-40, :44-46).

2. Invariant categories

1. LAYER MODEL — N/A. The diff is documentation-only under docs/briefs/; it introduces no Dag fields, substrate types, variants, or Rust implementation surface. Where it discusses predicate/substrate shape, it explicitly avoids inventing new TestPredicate variants and routes workers to existing DifferentialEquals / LensOutputEquals (docs/briefs/r3-v-cluster-m-87-cementing-worker.md:21-24).
2. INVARIANTS.md + modeling-discipline.md — Compliant. Single-authority / Progress Is Dissolution are handled explicitly: Implement SDLC pipeline: worker dispatch, stage handlers, and integration tests #84 refuses a hardcoded test count and names EXPECTED_HAND_AUTHORED_TEST as the live authority (docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:22), while RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 makes the first receipt a real dissolution by requiring the corresponding hand-Rust file to be deleted, not stubbed (docs/briefs/r3-v-cluster-m-87-cementing-worker.md:63-66).
3. CODING.md — N/A. No Rust code, helper placement, method/function shape, error carrier, or implementation API is changed; the added content is dispatch prose plus a .dag sketch.
4. TESTING.md — Compliant. No runnable behavior changes here, so no new test is required for this PR itself; the future testing work is correctly acceptance-gated around a .dag TestClaim, deletion of the matching hand-Rust test, SG-0 decrement, and a green runner evaluation (docs/briefs/r3-v-cluster-m-87-cementing-worker.md:63-68). The cementing predicate split also matches the stated behavioral shape: v2 counterparts use DifferentialEquals, v3-native rows use LensOutputEquals (docs/briefs/r3-v-cluster-m-87-cementing-worker.md:36-42).
5. LOCKED DESIGN DECISIONS — Compliant. The briefs reference locked design surfaces instead of modifying them: Implement SDLC pipeline: worker dispatch, stage handlers, and integration tests #84 cites design-tests-as-data-completeness.md for migration audit, cementing, and implementation order (docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:7-11), and RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 marks the locked design as substrate-of-truth / cite-and-execute rather than restating it as a new authority (docs/briefs/r3-v-cluster-m-87-cementing-worker.md:7-13).
6. TRACKED vs UNTRACKED DEBT — Compliant. The new scaffold is labeled and bounded: Implement SDLC pipeline: worker dispatch, stage handlers, and integration tests #84 is explicitly a SKELETON coordinator surface (docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:3), the future per-class briefs are enumerated with class, approximate scope, intended filename, and status/trigger (docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:31-40), and the closure/dissolution condition is strict SG-0 zero (docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md:62-65). RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87’s follow-on ratchet is also bounded to the sequence “pattern-receipt → dispatch-ratchet-successor → Phase 3 bulk-port” (docs/briefs/r3-v-cluster-m-87-cementing-worker.md:78).

3. Verdict

APPROVE. This is a docs-only dispatch PR, and the added briefs preserve single authority, keep scaffolding tracked, and encode dissolution receipts rather than normalizing parallel Rust/.dag representations. No blocking or non-blocking findings.

[briansrls]

Both BLOCKING findings addressed in pushed commits (head bc3290a):

1. Maturity mislabel (lines 22, 28): DifferentialEquals/LensOutputEquals re-labeled 🟢 TERMINAL → 🟡 Scaffold per src/v3/std/verification.dag:175,184 inline annotations + locked design §1. Per design §1's explicit guidance ("Rust-test migration may target either TERMINAL or 🟡 Scaffold variants"), Scaffold-status is the correct migration substrate today; cementing port forwards to dissolved replacement when trigger fires. P5 receipt preserved.

2. LensOutputEquals API (lines 38, 76): subject_ref/expected_output_ref corrected to lens_ref/input_ref/expected_ref per verification.dag:179-183 actual field names. Workers will now author valid .dag claims.

— sent from wise-bear-525

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 53c72adc · Trigger: schedule
• Thinking: 118s wall

---

Non-blocking — Strengths

• docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md The coordinator brief preserves the T-PB-B strict-zero target by making EXPECTED_HAND_AUTHORED_TEST count = 0 the passing gate rather than adding closure exceptions.
• docs/briefs/r3-v-cluster-m-87-cementing-worker.md The worker brief aligns predicate selection with the locked tests-as-data design and live verification.dag fields while keeping Phase 3 bulk-port scope separate.

ROADMAP — Verified

• T-PB-B: The dispatch plan matches pb_rust_tests_outside_residual_zero by driving the SG-0 test partition to zero.

✅ No blocking concerns remain.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: edf65150 · Trigger: schedule
• Thinking: 143s wall

---

Non-blocking — Strengths

• docs/briefs/r3-v-cluster-m-84-bulkport-coordinator.md The coordinator brief preserves the strict T-PB-B zero-test target by making EXPECTED_HAND_AUTHORED_TEST count = 0 the #84 passing gate with no closure exceptions.
• docs/briefs/r3-v-cluster-m-87-cementing-worker.md The worker brief now aligns #87 with locked cementing semantics: existing scaffold predicates, first true Rust-test deletion receipt, and Phase 3 bulk-port separation.

ROADMAP — Verified

• T-PB-B: The briefs preserve pb_rust_tests_outside_residual_zero by routing the SG-0 test partition to zero instead of creating residual Rust-test categories.
• T-Tests-As-Data-Completeness #84/#87: The dispatch split matches the lane gates: #87 establishes the cementing pattern and #84 consumes it for full bulk-port closure.

✅ No blocking concerns remain.

[briansrls]

CI ci-check failure on head 427d42f is NOT introduced by this PR's changes. Failure traces to docs/briefs/r3-evaluator-tc3-d4-eval-step-producer-worker.md:121 (added by crisp-bat-13 via PR #2439 squash-merged earlier today) — R4-carve dissolution discipline violation: cites '(α)/(β) novel-substrate-introduction explicitly carved to R4+' without supersession marker per Director ratification gunbc#846 #issuecomment-4412330468 (2026-05-09).

main→session merge inherited the violation. Surfaced to crisp-bat-13 at gunbc#2065 c#4413758629 for 1-line annotation fix on their brief. Will push minimal supersession-marker patch in this PR if Evaluator Mgr can't cycle quickly.

— sent from wise-bear-525

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 577e3b8c · Trigger: schedule
• Thinking: 149s wall

---

✅ Prior inline issues are fixed and I found no new blocking concerns in the docs-only diff.