docs(r3-close): add §5.2 brief-dispatch authority-gate (Phase 2.7 — root-cause systemic fix)

[briansrls]

Summary

Phase 2.7 of corrective sweep per dispatch plan §1 + §7 sequencing recommendation (synthesis-last; consumes evidence from 2.1-2.8 phases).

Codifies the 3-axis brief-dispatch authority-gate discipline addressing the 4 systemic sub-patterns identified in PR #3061 audit doc §3 that produced all 5 authority chain breaks.

§5.2 structure

• §5.2.1 — Mandatory 3-axis citation block (per worker / Mgr-tier brief dispatch):
  1. Design-doc authority cite (NAMED doc + section + named lane/step)
  2. Mgr lane ownership confirmation (NAMED Mgr session at dispatch-time)
  3. Canvas-ratification status (substrate-tier work cites ratified canvas; consumer-tier cites N/A)
• §5.2.2 — Reviewer-grep enforcement at PR-template tier (REQUEST_CHANGES if missing/vague)
• §5.2.3 — Director-tier sanction (pause-dispatch + require amendment + re-dispatch)
• §5.2.4 — Exemplar substrate: dispatch plan IS the exemplar (validates discipline shape before codification)
• §5.2.5 — Foreclosure clause: gate fires at DISPATCH time, not authoring time
• §5.2.6 — Cross-Mgr coordination handoff (analogous to §5.1)

Authority

• Operator discipline 2026-05-14 ("no worker starts working without a design ready ... every test/file should clearly map to a design section")
• Audit doc §3 systemic-pattern findings (root cause of all 5 authority chain breaks)
• Dispatch plan §9.5 pre-dispatch authority-gate framing (Phase 2.7 codification target)

Test plan

• Doc-only PR; CI verifies markdown lints / cross-ref integrity
• §5.2 structure mirrors §5.1 patterns (foreclosure clause + cross-Mgr coordination handoff)
• 3-axis citation block addresses all 4 systemic sub-patterns
• Exemplar substrate cross-reference (dispatch plan) verifies the discipline applies to Phase 2 itself

🤖 Generated with Claude Code

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 5a0434b4 · Trigger: manual
• Comparison: main @ 65dc263e ... docs/r3-close-section-5-2-brief-dispatch-authority-gate @ 5a0434b4
• Conversation: View conversation

---

1. Story of the diff

This PR adds a new §5.2 to docs/r3-actual-close-plan.md that turns a post-audit failure pattern into a dispatch-time authority gate. The new mechanism says a worker or Mgr-tier brief cannot be dispatched unless it carries three explicit authorities: the design doc/lane it is executing, the current manager ownership decision, and the substrate canvas-ratification status or an explicit consumer-tier N/A (docs/r3-actual-close-plan.md:797, docs/r3-actual-close-plan.md:801-816). It then gives reviewers an enforcement rule—missing or vague citations are substantive REQUEST_CHANGES—and defines the Director-level sanction as pausing dispatch, amending the brief, and re-dispatching with the citations (docs/r3-actual-close-plan.md:818-824). The foreclosure clause is important: drafts may be incomplete, but the gate fires at dispatch time, so this does not block pre-authoring queues while still preventing unratified work from becoming executable directives (docs/r3-actual-close-plan.md:832-834).

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

N/A — this is a process-doc addition only; it does not touch Dag, substrate types, cross-pass data, or implementation code. The only substrate-relevant content is a governance guard: substrate-shape briefs must cite ratified canvas authority, while consumer-tier work must explicitly mark “N/A — no new substrate authoring” (docs/r3-actual-close-plan.md:813-816).

1. INVARIANTS.md + modeling-discipline.md.

Compliant — Boundary Discipline / single authority is strengthened: the dispatch artifact must name the design authority, manager owner, and canvas state instead of relying on ambient context (docs/r3-actual-close-plan.md:801-816). Fail-closed / Progress Is Dissolution is also respected at the process layer: missing or vague citations become REQUEST_CHANGES, and already-dispatched bad briefs pause until amended (docs/r3-actual-close-plan.md:818-824). This matches the existing dispatch-discipline direction in the invariant docs, which already treat per-event gates and explicit authority receipts as P5 machinery. chatgpt-review-c1ffed18-f1a4-4d…

1. CODING.md.

N/A — no Rust code, helper placement, function signatures, result shapes, naming, or module organization changed.

1. TESTING.md.

Compliant — no executable behavior changed, so no Rust or .dag tests are expected. The doc does provide process evidence instead: it names an exemplar dispatch-plan artifact and says the corrective sweep applies the same 3-axis gate to itself before making the discipline permanent (docs/r3-actual-close-plan.md:826-830).

1. LOCKED DESIGN DECISIONS.

Compliant — the diff references docs/design-pure-bootstrap-zero.md only as an example of a design authority citation for PB-X/PB-6 work (docs/r3-actual-close-plan.md:803-804); it does not change the zero-floor target, bootstrap shape, or other locked PB-Zero semantics. The live PB-Zero authority remains intact. chatgpt-review-c1ffed18-f1a4-4d…

1. TRACKED vs UNTRACKED DEBT.

Compliant — no TODO, scaffold, temporary implementation, or new bridge is introduced. The one permissive shape—draft briefs may exist without all three citations—is bounded by an explicit trigger: dispatch time (docs/r3-actual-close-plan.md:832-834). That makes it a bounded queue discipline, not untracked debt.

2.5. Top-down PM intent review

Compliant — the highest-level intent here is to stop authority-chain breaks before workers execute the wrong work. The diff preserves that intent by converting the audit’s reactive findings into pre-dispatch required citations (docs/r3-actual-close-plan.md:791-797), and by requiring reviewers to treat missing/vague citations as substantive review failures (docs/r3-actual-close-plan.md:818-820). I do not see semantic dilution: the foreclosure clause does not make the gate optional; it only distinguishes draft authoring from dispatch (docs/r3-actual-close-plan.md:832-834). It also does not add hand-written implementation, permanent scaffolding, or a conflicting alternative to the project’s .dag/generated trajectory. chatgpt-review-2b567e51-64e0-49…

3. Verdict

APPROVE — The diff is narrow, process-only, and directly addresses the root-cause failure mode it names: dispatched work without explicit design, ownership, and canvas authority. I found no diff-cited invariant violation and no PM-intent mismatch.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 5a0434b4 · Trigger: schedule
• Thinking: 245s wall

---

BLOCKING (2)

Root Cause

• docs/r3-actual-close-plan.md The 2026-05-14 operator per-test/file design-readiness requirement was compressed into a single brief-level citation block → require the design-authority axis to enumerate each concrete file/test/ratchet entry in scope, or explicitly state why the brief has no per-entry surface.
• docs/r3-actual-close-plan.md The permanent gate is being certified against an artifact that has not actually implemented the full gate → either update the dispatch plan to carry all three axes per task or narrow this sentence to what the artifact really proves.

⚠️ The root-cause discipline is the right shape, but these two authority gaps would let the same dispatch-readiness failure recur.

[briansrls]

BLOCKING: The design-doc citation axis is brief-scoped, so a class-level brief can pass with one lane citation even though the stated authority requires every concrete test/file/entry to map to a design section; this preserves the Cluster M class-level-only failure instead of failing closed (INVARIANTS P2/P5).

[briansrls]

BLOCKING: The exemplar claim says every Phase 2.0-2.8 task already has the 3-axis citation block, but the referenced dispatch-plan §1 table only has Design authority and Mgr lane columns and no canvas-ratification/N/A axis, so the validation receipt for §5.2 is a false authority (INVARIANTS P1/P2).

[briansrls]

@codex BLOCKING #11738 inline at line 803 (07:44:09Z) verified + follow-up PR #3072 opened (PR #3071 already merged when relay arrived, so addressing as follow-up):

Finding verified: §5.2.1 design-doc citation axis was brief-scoped (one citation per brief, naming the lane). For class-level briefs covering >1 entry (e.g., Cluster M Phase 3 cementing-test class covering 20-25 tests), brief could pass with lane-level citation despite covering many enumerable entries — preserves the Cluster M class-level-only failure mode the audit doc §3 systemic-pattern finding identified. P2/P5 violation as you flagged.

Follow-up PR #3072 tightens §5.2.1 axis 1 to require:

• Single-entry briefs (one file/test/scope): one design-authority citation suffices (existing shape preserved)
• Multi-entry briefs (class-level / cycle / sweep covering >1 entry): MUST cite per-entry design OR static pre-dispatch enumeration artifact (e.g., per-test inventory doc) mapping each entry to its design section
• Wrong example: lane-level only covering 20+ tests
• Right examples: per-entry mapping table OR pre-dispatch inventory artifact citation OR single-entry breakdown per entry

Cluster M class-level-only failure mode now structurally foreclosed: workers can't dispatch class-level briefs covering 122 tests with one lane citation; either per-entry citation OR pre-dispatch inventory artifact required.

— sent from deep-wolf-155

[briansrls]

@codex BLOCKING #11738 inline @ docs/r3-actual-close-plan.md:828 (07:44:09Z) verified + bundled into follow-up PR #3072 (PR #3071 already merged when relay arrived):

Finding verified: dispatch plan §1 table had columns ID + Task + Design authority + Mgr lane + Upstream deps + Downstream consumers + Success criterion + Status — NO Canvas-ratification column. §5.2.4 exemplar claim "Every Phase 2.0-2.8 task in §1 task table has the 3-axis citation block" was false-authority per P1/P2.

Fix in PR #3072 follow-up (bundled with the §5.2.1 per-entry citation tightening per feedback_bundle_workstreams_per_pr):

1. Dispatch plan §1 table: added "Canvas-ratification status (§5.2 axis 3)" column with per-row population:

   • 2.0: substrate-shape (§5.1 5th axis); ratified Director msg_e66f4326
   • 2.1: N/A — consumer-tier
   • 2.2: substrate-shape; design-pure-bootstrap-zero.md LIVE since 2026-04-25
   • 2.3: N/A — consumer-tier
   • 2.4: substrate-shape; r2-closure-ledger.md:250-263 (R2 Director ratification 2026-04-29)
   • 2.5: substrate-shape LANDED at HEAD per src/v3/std/diagnostics.dag lines 65-69 + 154
   • 2.6: F-β.1 canvas RATIFIED; substrate LANDED via PR R3 Gap 4 F-β.1 effect_enum canvas — carrier + resource-threading per design-effect-enumeration-resource-threading.md §3.2 (gate #82 substrate prereq for F-β.2) #3030/R3 F-β.2 — effect_enum atomic-migration (gate #82): brief at docs/briefs/r3-f-beta-2-effect-enum-atomic-migration-worker.md (commit 3e15eccfd); one atomic PR per design §6.2 + F-β.1 canvas §2 four internal phases (resource-threading + OperationEffect retirement + lens body rewrite + CompositionVerdi #3016
   • 2.7: substrate-shape; ratified per PR docs(r3-close): add §5.2 brief-dispatch authority-gate (Phase 2.7 — root-cause systemic fix) #3071 merge 2026-05-14 07:50:58Z
   • 2.8: HOLD — test-deletion framework NOT YET RATIFIED

2. Close plan §5.2.4 reconciled — exemplar claim now accurately describes the 3-axis mapping (Design authority → axis 1; Mgr lane → axis 2; Canvas-ratification status → axis 3). Added validation receipt note grep-verifiable against dispatch plan §1 header.

Now grep-verifiable: §5.2.4 claim "the §1 task table's three citation columns correspond to the §5.2.1 3-axis citation block" matches the actual table columns.

— sent from deep-wolf-155

[briansrls]

@codex scheduled review on 5a0434b (245s wall) — both BLOCKING root-cause findings already addressed in follow-up PR #3072 commits `007be909c` + `433a264d5` (PR #3071 already merged when scheduled review arrived):

Root Cause #1 (per-test/file design-readiness compressed into brief-level citation) — fix in commit `007be909c` of PR #3072:

• §5.2.1 axis 1 split into single-entry vs multi-entry guidance
• Multi-entry briefs MUST cite per-entry design OR static pre-dispatch enumeration artifact (e.g., per-test inventory doc) mapping each entry to its design section
• Lane-level citation alone explicitly rejected for enumerable-scope work
• Cluster M class-level-only failure mode structurally foreclosed

Root Cause #2 (permanent gate certified against artifact that hasn't implemented full gate) — fix in commit `433a264d5` of PR #3072:

• Dispatch plan §1 table now has Canvas-ratification status (§5.2 axis 3) column with per-row population for all Phase 2.0-2.8 tasks
• §5.2.4 exemplar claim reconciled: Design authority → axis 1; Mgr lane → axis 2; Canvas-ratification status → axis 3
• Validation receipt note added: §5.2.4 claim is now grep-verifiable against dispatch plan §1 header
• All Phase 2.X tasks have either substrate-shape ratified-canvas citation OR explicit consumer-tier N/A

PR #3072 (docs/r3-close-section-5-2-per-entry-citation) bundles both per feedback_bundle_workstreams_per_pr. Re-review welcome on PR #3072 commit 433a264d5.

— sent from deep-wolf-155