Elnor Repo Reader

# Stage 5R3 Pass 1 — Architect Decisions Resolved

**Date resolved:** 2026-05-28
**Resolved by:** Will (via Cowork conversation 2026-05-28 with Claude)
**Companion source:** `PASS_1_ARCHITECT_DECISIONS.md` (the open decision queue Pass 1 produced)
**For consumption by:** Pass 2 — read this file as authoritative architect guidance. Where this file gives explicit Pass 2 instructions, follow them verbatim. Where this file defers something, do NOT promote it back into scope.

---

## Decision 1 — Pass 2 gate (Stage 5R2 regression review)

**Decision:** **Procedural — pending external review.**

Will routed the regression review prompt to red-team. Pass 2 cannot start until the response lands in `Memory Rebuild Docs/Flattening/Reviews/Red Team Responses/Stage_5R2_Regression_Review_Response.md` AND ratifies Stage 5R2 stabilization.

**Pass 2 instructions:**
- Check for the response file at session start. If absent, STOP and surface in `PASS_2_WILL_REVIEW_PACKET.md`.
- If response forces edits to any of `DOC80 Skeletal Baseline`, `Owner Map`, `Import Graph`, `Retired Names`, refresh `baseline_snapshot/` BEFORE running any retargeting logic. Re-verify sha256 against live sources.
- If response ratifies without edits, proceed.

---

## Decision 2 — ADQ-222 (Phase-1 sharing schemas)

**Decision:** **Tracked seam only. Do NOT draft `PublishedViewEnvelope`, `TaskSharedMemoryExposureContract`, or `PublishedLibraryCorpusExposureContract` schema bodies.**

Sharing is forward-looking; drafting schema bodies now during a major memory flatten creates premature commitment risk. ADQ-222 explicitly does not block Stage 6 (per RUN_STATE.md).

**Pass 2 instructions:**
- Leave ADQ-222 in the ADQ ledger as `open / tracked_seam_only`.
- Do not generate OP-A V4 rows for the three sharing schema bodies.
- Do not retarget any current row to a sharing schema body destination.
- One narrative line in `STAGE_6_CHARTER_INPUT_INDEX.md` noting "sharing schemas deferred per ADQ-222 resolution Stage 5R3."

---

## Decision 3 — 38 missing explicit OBL-* IDs

**Decision:** **Split treatment.** DOC24 R3.1.1 IDs get added now; DOC23 Addenda B IDs are deferred entirely.

### 3a — DOC24 R3.1.1 (6 IDs absent from OP-A)

**Add as new OP-A V4 rows. No further architect review.**

**Pass 2 instructions:**
- For each of the 6 IDs identified in `PASS_1_INVENTORY.md` Self-Audit Coverage Addendum and `PASS_1_SELF_AUDIT.md §28`, add a new OP-A V4 row in the appropriate target-doc §6 section.
- Field defaults: `audit_pass=1`, `confidence=high`, `architect_review_required=false`, `flatten_status` per row content (Pass 2's normal classification logic applies), `disposition` per Pass 2 retargeting logic.
- Source: `Current Specs/DOC24/DOC24_R3_1_1.md`.
- These go directly into **Bucket A** (confident retarget); no Bucket C diversion.

### 3b — DOC23 Addenda B (33 IDs across 3 files)

**Defer entirely. Do NOT add as OP-A V4 rows.**

Will is actively working on DOC23 Addenda B (Outcome Evaluator/Revisor V3.3.1 + Source Workspace V1.0.1 + Task Forum Run Board V1.0.1) and has deliberately held off cross-doc obligation registration — particularly the learning-substrate obligations — until the memory flatten settles.

**Pass 2 instructions:**
- Do NOT generate OP-A V4 rows for the 33 deferred IDs.
- Add a single tracking note in `OPA_V4_CANDIDATE.md §9` (Open Meta-Architecture Questions): **"DOC23 Addenda B cross-doc obligation registration deferred per architect 2026-05-28; 33 explicit OBL-* IDs in Outcome Evaluator/Revisor V3.3.1 (29), Source Workspace V1.0.1 (3), Task Forum Run Board V1.0.1 (1) are held pending DOC23 Addenda B stabilization post-flatten. Re-evaluate at OP-A V5 or when DOC85 lands (whichever first)."**
- Add a tracking row in `DOC73_V1_6_DEFERRAL_INVENTORY_R1.1.md` parallel mechanism (NEW row `DOC23_ADDB_OBLIGATION_REGISTRATION_DEFERRED`) — Pass 2 does this as a side artifact.
- These 33 IDs do NOT go to any bucket. They are explicitly out of Stage 5R3 scope.

---

## Decision 4 — S9-TRACKING-SURFACES (collapse vs. per-doc companions)

**Decision:** **Collapse to OP-A as the single active surface.**

Per-doc companion trackers cause drift (well-documented pattern). One tracker, updated at session close, beats N trackers that fall out of sync.

**Pass 2 instructions:**
- In rewritten §15A, document: "OP-A is the single active cross-doc obligation surface for ELNOR. Per-doc companion trackers are deprecated. Specs MAY include §22-style cross-doc obligation sections inline for spec-internal navigation, but those sections must be reflected as rows in OP-A — they are NOT a parallel truth surface."
- In §5 source-rule refinement, clarify that "companion tracker" sources are out-of-scope for §3 going forward.
- Close §9 line 7430 with this resolution.

---

## Decision 5 — S9-CD-MASTER-INHERITED (inherited tracker inventory)

**Decision:** **Close as no-action. Mark §9 item resolved.**

The inherited CD Master Integration Index R1 tracker list predates OP-A's promotion to standing instrument. OP-A subsumed its function.

**Pass 2 instructions:**
- Close §9 line 7431 with: "RESOLVED 2026-05-28: CD Master Integration Index R1 tracker inventory subsumed by OP-A as single active surface (per Decision 4). No individual tracker from that index needs separate §3 source-registry treatment."
- No rows added; no rows changed.

---

## Decision 6 — S9-MEMORY-INTAKE (Memory Intake and At-Use Disciplines proposal)

**Decision:** **Defer. Do NOT adopt during Stage 5R3.**

Whatever this proposal addressed is being substantially reworked by the DOC80-87 family redesign. Adopting it mid-flatten risks contradictions. Revisit after V4 ships.

**Pass 2 instructions:**
- Close §9 line 7432 with: "DEFERRED 2026-05-28: Memory Intake and At-Use Disciplines proposal not adopted during Stage 5R3. Re-evaluate post-V4 against DOC80-87 schemas; if specific concerns are still real, fold into a DOC85 or DOC83 addendum at that time."
- Do NOT add rows for proposal content.
- Add to `DOC73_V1_6_DEFERRAL_INVENTORY_R1.1.md` as `MEMORY_INTAKE_AT_USE_DISCIPLINES_POST_V4_REVIEW`.

---

## Decision 7 — S9-DOC74

**Decision:** **Typo / no-action. DOC74 does not exist; reference is a numbering artifact.**

DOC73 is followed by DOC80 in the spec numbering (DOC74-79 unused — the gap was intentional headroom). Any reference to "DOC74" in OP-A or §9 is a numbering confusion.

**Pass 2 instructions:**
- Close §9 line 7435 with: "RESOLVED 2026-05-28: DOC74 reference is a numbering artifact. No such doc exists; no such doc planned. No action required."
- Do NOT create any DOC74-targeted phantom row.

---

## Decision 8 — S9-D14-META (OBL-D14-09 walking D15-RT-001..005)

**Decision:** **In-scope. Inventory as live Stage 5R3 retarget candidates.**

These are row-shaped obligations sitting in OP-A; Stage 5R3 is the right place to walk them.

**Pass 2 instructions:**
- Include `OBL-D14-09` and `OBL-D15-RT-001` through `OBL-D15-RT-005` in the retargeting walk.
- Each row gets the standard tagging (`pre_flatten` / `flatten_aware` / `not_memory`) per Task A.
- Retarget per Owner Map per Pass 2 logic; no special handling beyond standard flow.
- Close §9 line 7439 with: "RESOLVED 2026-05-28: meta-obligation chain in-scope for Stage 5R3 retarget."

---

## Decision 9 — S9-RUNNING-BRIEF (Running Brief slot disposition)

**Decision:** **DEPRECATED-with-capture.** The Running Brief slot itself is gone; valuable concepts already ported into DOC73 V1.6 + planned DOC80-87 destinations per the 2026-05-04 CSA extraction.

Will reviewed CSA R2 vs DOC73 V1.6 vs DOC80 Skeletal Baseline 2026-05-28 and confirmed:
- RecentActivityRollup (the core CSA data structure) is already in DOC73 V1.6 Artifact 1 §16.2.
- RecentActivityRollup consumption contract is planned at DOC83 §2.4 per ADQ-405.
- Session-orientation injection (CSA's "fresh-surface" idea) maps to DOC84 §9-11 (CarryoverCapsule + CognitiveDiff + Resume card + pipeline park/resume).
- Specific Tier 1/2/3 tiering policy was deliberately deleted (CSAInjectionTierPolicy / `inject_csa_orientation` / `OrientationContextEntry` / `tier_strategy`).
- Architect preference for `tier_3_minimal_orientation` semantics preserved in `BUILD_QUESTIONS §11.1` as a future-design hint.
- Runtime tiering work tracked as deferred row `DOC73_TO_DOC72_CSA_INTERFACE_REVIEW` in `DOC73_V1_6_DEFERRAL_INVENTORY_R1.1.md §1.2`.

**Pass 2 instructions:**

1. **Apply OP-A row changes flagged in `DOC73_V1_6_CSA_EXTRACTION_REPORT.md §6 verbatim:**
   - **REMOVE 3 rows (per §6.1):**
     - `OBL-D72-CSA-R2-DOC73-ALIGN-01` (OPA V3.18 line 3285 — DOC73 V1.6 no longer absorbs CSA; obligation has no consumer)
     - `OBL-D72-CSA-R2-MECH4` (OPA V3.18 line 3297 — redundant with `OBL-D73-V16-MECHANISM4-01`)
     - `OBL-D73-CSA-R2-ABSORPTION` (OPA V3.18 line 3375 — no longer relevant post-extraction)
   - **RENAME/RESCOPE 2 rows (per §6.2):**
     - `OBL-D73-N-ORIENTATION-INV-01` → `OBL-D73-N-NOT-EVIDENCE-INV-01` (preserve underlying "rollup not evidence" invariant per `INV-N-NOT-EVIDENCE-1`)
     - `OBL-D73-V16-MECHANISM4-01` description rewrite — strip "CSA recent_activity may orient query framing"; replace with "RecentActivityRollup cannot satisfy legal evidence queries per INV-N-NOT-EVIDENCE-1"
   - **ADD 1 row (per §6.3):**
     - `OBL-D73-RECENT-ACTIVITY-ROLLUP-CONSUMER-CONTRACT-01` (canonical home Artifact 1 R0.5 §16.6; retargets to **DOC83 §2.4** in V4 per ADQ-405)

2. **Retarget any remaining "Running Brief"-targeted DOC24/DOC15 rows** per this map:
   - Schema/producer obligations → DOC73 (already ported)
   - Consumption/orchestration obligations → DOC83 §2.4
   - Session-start injection / tier-1 continuity card semantics → DOC84 §9-11
   - Specific Tier 1/2/3 tiering policy obligations → **DEFER** (do not retarget; mark as deferred per `DOC73_TO_DOC72_CSA_INTERFACE_REVIEW`)

3. **Close §9 line 7440** with: "RESOLVED 2026-05-28: Running Brief slot = DEPRECATED-with-capture. CSA R2 concepts already extracted into DOC73 V1.6 (2026-05-04); destinations mapped to DOC73/DOC83/DOC84 per DOC80 Skeletal Baseline §2. Runtime tiering deferred via existing deferral row."

4. **Architect preference preservation:** `tier_3_minimal_orientation` preference per `BUILD_QUESTIONS §11.1` carries forward as a future-design hint for DOC84 charters. Pass 2 does NOT need to encode it in V4; it's already preserved in DOC73 build records.

---

## Decision 10 — S9-DUPLICATE-PAIRS (V3.8 duplicate-row pairs)

**Decision:** **Punt to Pass 2 with merge default.** For each duplicate pair, merge with the higher-`confidence` row winning; loser becomes alias.

**Pass 2 instructions:**
- For each duplicate pair (2), (3), (4) flagged in OPA V3.18 §9 line 7449:
  - Compare the two rows' `confidence` field.
  - Merge: winner keeps row ID + canonical content; loser ID goes into winner's `legacy_id_aliases` metadata field for traceability.
  - Tie-breaker: if `confidence` is equal, prefer the row whose `source_doc` is currently operative; if still tied, prefer lower row_id.
  - **Escalate to Bucket C ONLY if** `confidence` is equal AND `source_doc` for the two rows are different AND `new_target_doc_or_docs` differ post-retarget.
- Close §9 line 7449 with: "RESOLVED 2026-05-28: duplicate-pair merge default applied during Pass 2."

---

## Decision 11 — S9-V161-LANE + S9-V18-ID-NORMALIZATION

**Decision (11a — V1.6.1 lane scope):** **In-scope for Stage 5R3 retarget.**

DOC73 V1.6 is part of the active memory wave being restructured by the flatten.

**Decision (11b — V1.7→V1.8+ ID normalization):** **Normalize to V4 single-row shape.**

Old V1.7 IDs become aliases in row metadata for traceability. Carrying two parallel IDs forever is exactly the registry-drift to avoid.

**Pass 2 instructions:**
- V1.6.1 rapid-patch-lane rows included in retarget walk; standard tagging applies.
- When Pass 2 hits a row that V3 tagged with a V1.7 candidate ID, normalize to the V4 single canonical row. Old V1.7 ID → `legacy_id_aliases` metadata field on the V4 row.
- Close §9 lines 7450 + 7451 with: "RESOLVED 2026-05-28: V1.6.1 in-scope; V1.7 IDs normalize to V4 single-row shape with legacy_id_aliases preserved."

---

## Decision 12 — S9-INJECTION-SLOTS (Core R0.8 InjectionSlotRegistry gap)

**Decision:** **Owner-doc work; not a Stage 5R3 prerequisite.**

Pass 2 surfaces the gap as an obligation row pointing at Core R0.8; the actual slot registry mechanics get drafted when Core R0.8 sees its next revision.

**Pass 2 instructions:**
- Surface as a single obligation row in OP-A V4 targeting `DOC23 Addenda B Core` (not memory-plane).
- Tag as `not_memory` per Task A; do NOT retarget to DOC80-87.
- Disposition: `pending — fold on Core R0.8 next revision`.
- Close §9 line 7453 with: "RESOLVED 2026-05-28: InjectionSlotRegistry gap remains owner-doc work; surfaced as OP-A row only."

---

## Decision 13 — S9-DOC25-DEGRADED + S9-KNOWLEDGE-PACK + S9-NIGHTLY-EXTRACTION

### 13a — DOC25 quality/degraded-state reporting (§24.3.6)

**Decision:** **Add explicit DOC25 surface row.**

Partial coverage in IngestionResult acceptance is a real obligation hole; documenting it gives Pass 2 a target. Post-flatten retargeting likely lands the row in DOC85 (learning) or DOC83 (extraction quality), depending on Pass 2's Owner Map walk.

**Pass 2 instructions:**
- Add new OP-A V4 row: `OBL-D25-DEGRADED-STATE-REPORTING-01` targeting DOC25 §24.3.6.
- Pass 2 retargets per Owner Map — likely lands in DOC83 (extraction spine) or DOC85 (learning), but Pass 2 decides.
- `audit_pass=1`, `confidence=medium`, `architect_review_required=false` (Will pre-approved the row addition).
- Close §9 line 7454.

### 13b — EC Core Knowledge Pack compiler jobs (§24.6.2)

**Decision:** **Target DOC85 (Learning).**

Per DOC80 Skeletal Baseline §2: DOC85 owns learning architecture + write-back path + warrant-degradation-trigger producers. Knowledge Pack compiler jobs are learning-substrate work.

**Pass 2 instructions:**
- Knowledge Pack compiler job obligations retarget to **DOC85**.
- Loses pre-flatten EC Core ownership; new owner is DOC85.
- `confidence=medium` (DOC85 is in design, but the owner boundary is unambiguous per ADQ-221 and DOC85 §1.1 rescope at Patch B6).
- Close §9 line 7455.

### 13c — EC Core nightly/ongoing extraction jobs (§24.6.6)

**Decision:** **Target DOC85 (Learning).**

Same reasoning as 13b. Nightly extraction is learning-substrate work, not orchestration. EC Core retains execution-of-write per DOC85 §1.2 ("EC executes the durable write"), but the ownership of the extraction job semantics lives in DOC85.

**Pass 2 instructions:**
- Nightly extraction job obligations retarget to **DOC85**.
- Loses pre-flatten EC Core ownership; new owner is DOC85.
- `confidence=medium`.
- Close §9 line 7456.

---

## Items NOT requiring Pass 2 action (already resolved or carry-forward)

For completeness, the following OP-A §9 items were already resolved or clarified in prior OP-A revisions and require no Pass 2 action:

- Line 7433: DOC24↔DOC73 full seam audit — RESOLVED in V3.7
- Line 7434: DOC6 status — RESOLVED in V3
- Line 7436: DOC15 R3.1 contract maintenance — RESOLVED in V3
- Line 7437: DOC1 R1 status — ordinary future review (not Stage 5R3)
- Line 7438: DOC15-derived row triage — exactly what Pass 1/2 is doing (no separate decision)
- Line 7441: EC Core agent registry ownership — RESOLVED in V3.5 (architect Option A)
- Line 7442: Agent-vs-Capability registry split — CLARIFIED in V3.6
- Line 7443: Scope vocabulary reconciliation — existing DOC24/DOC1 item
- Line 7444: RRB-derived row triage — covered by Pass 1/2 workflow
- Line 7445: DOC2 retirement coordination — likely owner-transfer; Pass 2 inventories
- Line 7446: DOC24↔DOC73 V1.5.1 audit complete — RESOLVED in V3.7
- Line 7447: V3.8 patch session as V4 §0.2.1 process gate — process-gate history
- Line 7448: Two emitter/consumer dedup pairs — pair #2 resolved by V3.11; pair #1 covered by Decision 10 above
- Line 7452: DOC24 R3.1.1 V1.x audit complete — RESOLVED in V3.9
- Line 7458: V5.2 sub-agent fold-in — RESOLVED in V3.16

---

## Pass 2 acknowledgment checklist (Pass 2 reads this section before starting)

Before Pass 2 begins retargeting work, confirm:

- [ ] Stage 5R2 regression review response present at `Memory Rebuild Docs/Flattening/Reviews/Red Team Responses/Stage_5R2_Regression_Review_Response.md` (Decision 1). STOP if absent.
- [ ] Baseline snapshot refresh applied if regression review forced edits to any baseline file (Decision 1).
- [ ] sha256 re-verified against live sources for all baseline_snapshot files.
- [ ] This file read and acknowledged.

Acknowledgment: by proceeding past this checklist, Pass 2 commits to following the resolutions above verbatim. Where this file is silent, Pass 2 falls back to the prompt at Appendix B.3 of `Stage_5R3_Audit_Proposal_R3_1.md`.