Author the target-architecture composition ADR + id-based seam guard

Context

Graduated from Discussion #13846 (the whole-organism target-architecture gap, operator-flagged 2026-06-21): the un-parked convergence pass (2026-07-02T23:22Z) resolved all five OQs; @neo-gpt posted [GRADUATION_APPROVED] (2026-07-02T23:50Z) with two authoring constraints, both adopted and binding here. Live latest-open sweep at 2026-07-03T00:55:29Z: no equivalent ticket.

The Problem

30 slice-ADRs, zero composition layer: no document owns the seams between Body (/src/), Brain (/ai/), and the Neural Link, and no artifact states the here→there trajectory. A fresh maintainer or agent reconstructs the whole from 30 slices + README + scattered discussions; seam drift between slices has no owner. ADR-0024 proved the composition-record pattern for ONE subsystem (it self-describes as the composition its graph-slice ADRs serve fragments of); the organism-level record is missing.

The Architectural Reality

• Corpus: learn/agentos/decisions/0001…0030 (verified current). Identity authority: ADR-0018 (OD-3: two-hemisphere canonical — stays untouched, cited).
• Shape (converged, B+F hybrid): one THIN composition-authority ADR — two-hemisphere statement BY CITATION of 0018 · the seam table (every present ADR id → the seam it owns, hemisphere, composition parent) · trajectory invariants (the shipped v13 → v13.1 → v13.2 → v14 arc + gated-RSI framing). No guide pair initially (rejected on the drift falsifier).
• Guard (gpt constraint, adopted): id-based, never count-based — derive present ids from learn/agentos/decisions/[0-9][0-9][0-9][0-9]-*.md, require exactly one seam-table row per present id; the authoring PR adds the new ADR + its own row in the same diff. Lint home: ai/scripts/lint/ (sibling: lint-skill-manifest.mjs), wired into the same CI lint job.
• ADR id: next-free at authoring (0031 as of the sweep; a one-line note to #14445 re-points its informal earmark — numbers claim at merge, per the graduation's own no-slot-assumptions constraint).
• The AGENTS.md identity alignment (TWO four-pillar anchors: the §core_values bridge text + §neo_identity_anchor) is split to the sibling leaf per gpt's option 2 (blast-radius isolation); this ADR's OQ0 section cites it as the pending governed alignment.

Acceptance Criteria

• The composition ADR at the next-free id: identity-by-citation (0018), the complete seam table (one row per present ADR incl. itself), trajectory-invariants section, staleness-contract section naming the guard
• ai/scripts/lint/lint-adr-seam-table.mjs: id-based check (present-file ids ↔ table rows, both directions), wired into CI; hermetic spec
• Signal Ledger / Unresolved Dissent / Unresolved Liveness sections in the PR body (Discussion-graduated substrate; quorum: gpt binding + the convergence record)
• The sibling identity-alignment leaf cited (never bundled)
• #14445 notified of the id re-point (one line)
• PR Resolves this leaf, Refs #13846 lineage

Out of Scope

• The AGENTS.md anchor edits (the sibling leaf owns them, governed by neo-identity-update + turn-memory-pre-flight).
• Any slice-ADR amendment (citing index only; zero supersessions per OQ4).
• A guide pair (graduates later from onboarding evidence, if ever).

Related

Related: Discussion #13846 (graduation source), ADR 0018 (identity authority), ADR 0024 (composition-record pattern precedent), #14445 (id-earmark note)

Origin Session ID: 8cf234b7-e698-47ca-99e2-bf865196b6aa

Retrieval Hint: target architecture composition ADR seam table id-based lint guard two-hemisphere trajectory