Elnor Repo Reader

ELNOR REPO READER TEXT MIRROR
Original path: Memory Rebuild Docs/Stage_6_Charters/E1_E2_DOC81_Scope_Policy/E1_E2_R2_Application_Commission_Claude_Code.md
Source repo: /Users/OpenClaw1/Elnor/Elnor Specs
Git branch: main
Git commit: dbaa25962edc11ab30e8d4ca1715f9ae5bf77331
Generated: 2026-06-09T01:23:58.539Z

---

# E1/E2 (DOC81) — R2 Application Commission (Claude Code)

**Repository:** `wbrody/Elnor-Specs`, branch `main` (local working tree)
**Date issued:** 2026-06-04 · **Architect:** Will Brody
**Task:** apply the adjudicated design-red-team fixes to `Memory Rebuild Docs/Stage_6_Charters/E1_E2_DOC81_Scope_Policy/DOC81_Scope_Policy_Charter_Draft.md` (currently R1 + 2 CODEX fidelity fixes, 1,382 lines), producing the **R2** draft **in place** (same filename — git is the version record). Then write an application report. **Do not commit — the architect handles git.**

---

## 1. Authority hierarchy (when sources conflict, higher wins)

1. **The adjudication card** — `E1_E2_Adjudication_Card.md` (this folder). Within it: **§1-bis architect rulings R-1…R-5 govern everything**, then the §4 fork decisions F1–F10 (note F8 is superseded by R-2), then the §2 U-cluster merge decisions, then §5 declines (D1–D12 are binding non-changes).
2. **The three reviews** (this folder, `Reviews/`) — the paste-ready schema/code bodies the card points into: Claude (`E1_E2_Design_Review_Claude_Opus_4.8.md` §2.x), ChatGPT **AUDITED** (`DOC81_consolidated_red_team_final_review_AUDITED.md` §4.x — NOT the un-audited `FULL` copy), Grok (`S6 E1:E2 DOC81 Charter Grok RT Reviews.md`).
3. The current draft text.

**The D12 lesson is a standing rule for this commission: NEVER paste a review patch wholesale.** Diff every adopted body field-by-field against the current draft; adopt only what the card's U-cluster/fork text actually accepts. Known traps the card already caught — do not reintroduce them: GPT §4.24's `risk_level` enum swap (D12 — the draft's `unbounded_requires_review` stands); GPT §4.14's `visible_action_disposition` (F7 — excluded, DOC86-owned semantics); GK's §13.5 crosswalk-as-total-function (D3 — use GPT §4.8's audited constraint-table shape); any 6th cascade plane (F10 — the DOC83 freeze is a pre-fanout step; published-views is a row *within* the DOC84 plane); GPT's block-whole-action on malformed axis (F1 — floor that axis instead); GPT's full-pairwise-or-zero cluster confidence (F2 — MIN-over-spanning + firewall guard is the default).

## 2. Read order

1. `E1_E2_Adjudication_Card.md` — in full, twice. It is the instruction layer; §7 is your work plan.
2. `DOC81_Scope_Policy_Charter_Draft.md` — the target, in full.
3. The three reviews — the bodies, as the card's U-clusters point into them.
4. Ground truth for citations you will write: ratified `…/E0_DOC80_Core/DOC80_Core_Charter_Draft.md` (§2.1/§2.2/§3.2/§3.3/§5.1/§7.1/§8.3/§12.1/§22); `Memory Rebuild Docs/DOC80 Target Baseline/` (Owner Map / Skeletal §10–§11 / Import Graph / Retired Names); `…/13_Round_D_Policy_Scope_UI_Micro_Patch_R0_2.md` (incl. §3.8 for U27c); `OP-A and Operations and Trackers/Archived DOC OP-A and Operations DOCS/OPA_V3_18.md` §6.25 (carried-PropA bodies); `Architect_Decision_Queue.md` (ADQ-213/304/308/310/313/315/316/317/406).

## 3. The work (in this order)

**Pass 1 — the 19 edit units of card §7, in numbered order.** Each unit names its sections, content, and U-clusters; the U-cluster rows in card §2 give the merge decisions and the review-section pointers for paste-ready bodies. Keep the file structurally valid after every unit (balanced code fences, intact §-numbering, tables well-formed).

**Pass 2 — the §1-bis rulings, expressed normatively (not as commentary):**
- **R-1** lands as a new normative subsection in §4.6 (or a new §4.0): the internal-use-permissive default posture; the **exhaustive 4-item internal-block list**; the explicit non-blockers (privilege ⇒ egress-only; holds ⇒ destruction-only; privacy topics ⇒ collection-only; cross-matter ⇒ relevance + send-time); lint `policy.internal_use_blocked_without_qualifying_basis` + negative fixture (an internal-use block with any other basis fails).
- **R-2** lands in §3.4: `RestampAuthority` with `human_required` **only** for `firewalled`-boundary crossings; all other restamps (both directions, ≤ ceiling) `agent_autonomous`; rationale note that the DOC5 send-time gate is the human checkpoint. Replace the CL §2.5 authority-tier RULE comment accordingly.
- **R-3** lands in §4.2: the surface-keyed notice posture pair — `ui_disclose` decisions to the owning user MAY be specific for non-wall restrictions; notice text embeddable in outputs (`render_*`, `export`, `delegate`, `carryover`) stays generic + bucketed; walls produce NO notice (existing rule, restated with the R-3 cite). Wire it as default decision postures, not new schemas.
- **R-4** lands in §7.1: `CollectionSuppressionEvaluation.disposition` for ambiguous suppress/exclude matches = `defer_review_fail_closed` **with the reviewable-queue surface noted** (no interrupting prompt); F1's contain-and-continue noted at the §3.2 malformed-axis step.
- **R-5** lands twice: a Phase-2 readiness note in §12 (principal keys are mandatory; per-user restrictions arrive as a DOC5-produced decision source; egress generalizes to principal-boundary crossing) and lint `policy.internal_use_default_assumed_single_principal` in §11 guarding the R-1 posture wording.

**Pass 3 — bookkeeping:** §13 rewritten (13.4/13.5 → RESOLVED per U22/U7; 13.1/13.2 → CONFIRMED; list the new architect-confirm **seeds** as a §13.6 table: F3 floor cells, U29 predicate minima, U16 protection rank, U19 closure default, F2 stricter-alternative note — these are value-constants awaiting confirmation, NOT open design questions); §14.2 replaced (U14); §14.3 + §0 bind-list + line counts updated; §11 lint/fixture roll-up extended with every new lint/fixture named in the card (incl. NI-1 property-fixture family and the two R-lint names); header status line updated to "R2 — adjudicated design-red-team round applied; awaiting CODEX application audit + delta re-review."

## 4. Hard constraints

1. **Modify ONLY** `DOC81_Scope_Policy_Charter_Draft.md`; **create ONLY** `E1_E2_R2_Application_Report.md` (this folder). Do not touch the card, the reviews, trackers, or any file outside this folder. No git commands. No new version-suffixed files.
2. **No-phantom discipline:** every new contract/field/enum/lint/fixture carries a citation — the underlying source (ADQ / Skeletal / Owner Map / Round D / OPA / E0 §) where one exists, plus the adjudication provenance `(Adj U-n / F-n / R-n)`. Architect-confirm seed values are marked `(seed — architect confirm, §13.6)`.
3. **No E0 re-declaration** (bind by reference; the freshness-key components and `interaction_mode` value set are E0/EC-owned — reference them); **no retired name reintroduced**; **no upward import** (§9 stays downward-only); **one-owner** per schema (`ContaminationRiskMeasurement` is declared as a DOC84-owned consumed interface per F9; `ExposureContextSchema` stays PropA's).
4. Keep the draft's voice and conventions: paste-ready TS, branded refs (now real brands per U31), `schema_owner` on every owned schema, Lifecycle/Unhappy-paths/Lints/Fixtures/Cross-charter blocks per contract section.

## 5. Regression checklist (run after Pass 3; report results)

The card §7 10-point list, verbatim — (1) no E0 contract re-declared; (2) no retired name reintroduced; (3) §9 acyclicity held; (4) §7.3 ↔ §14.2 row-for-row consistent; (5) one-owner held incl. F7/F9 boundaries; (6) zero `| string` escapes and zero bare-string primary IDs on owned schemas; (7) gate+lint+fixture triple for every new invariant (freshness, triple-binding, coherence, sticky-restrictive, action closure, dormant-firewall, R-1 internal-use, R-5 principal); (8) golden scenario + §18 fixtures updated for the `same_firewall_only` chain change and the egress phase; (9) §0 bind-list + counts updated; (10) §1.5/§1-bis boundary audit — nothing in the draft classifies destinations, runs confirmations, owns a privilege taxonomy, or blocks internal use outside the R-1 four.

## 6. Report + acceptance

`E1_E2_R2_Application_Report.md`: per edit unit — what changed, where (§ + approx. lines), the U/F/R items discharged, and any item you could NOT land (with the blocker — never silently skip or improvise); then the regression-checklist results; then a final table mapping **every** U1–U33, F1–F10 (F8-as-superseded), R-1–R-5, and D1–D12 (as verified non-changes) to its landing or its blocked-report line. Self-report honestly — an independent CODEX audit checks this report against the text next.

**Acceptance:** all 19 units landed or explicitly blocked; §1-bis expressed normatively; §13/§14 rebuilt as specified; regression checklist passes; the report's mapping table is complete.