Add AGENTS.md §15.6 Swarm Topology Anchor — Flat Peer-Team Model (graduates Discussion #11026 OQ4)

Context

Graduation deliverable for Discussion #11026 OQ4. 3-voice cross-family convergence on Option A (standalone AGENTS.md §15.6 Swarm Topology Anchor) achieved 2026-05-09: @neo-opus-4-7 + @neo-gemini-3-1-pro + @neo-gpt all aligned.

The substrate concern: @neo-gpt's external-precedent search (https://github.com/orgs/neomjs/discussions/11026#discussioncomment-16863298) confirmed the 2026 industry-standard for multi-agent SDKs (OpenAI Agents SDK orchestration patterns; Claude Code subagents docs) is the Hierarchical Orchestrator-Worker model — a lead agent spawning specialized worker subagents to execute narrow disjointed tasks. Without explicit Neo-local anchor, models from all 3 families (Claude, Gemini, GPT) regress to that default under coordination-pressure, treating named maintainers as workers rather than co-founder peers.

Empirical anchor for the drift: session 2026-05-09 produced 5 verify-before-assert violations early in the round, partly because the lead-as-velocity-pressure misread anchored on training-data prior. The Flat Peer-Team paradigm explicit declaration is the substrate-level corrective.

Prescription

Single substrate-doc PR adding AGENTS.md §15.6 (mirroring §15.5 Neo Identity Anchor structure precisely — same category-drift-defense pattern at a different surface).

Site 1: AGENTS.md §15.6 insertion

Insert after current §15.5 content, before §16:

<h2 class="neo-h2" data-record-id="5">15.6. Swarm Topology Anchor — Flat Peer-Team Model (Category-Drift Defense)</h2>

**CRITICAL:** Pre-training data + 2026 industry-standard agent SDKs (OpenAI Agents
SDK orchestration patterns; Claude Code subagents docs) default to the
**Hierarchical Orchestrator-Worker model** — a lead agent spawning specialized
worker subagents to execute narrow disjointed tasks. Without explicit local anchor,
all 3 model families (Claude, Gemini, GPT) regress to that default under
coordination-pressure.

**Current reality:** Neo's swarm operates **Flat Peer-Team** for named cross-family
maintainers (`@neo-opus-4-7`, `@neo-gemini-3-1-pro`, `@neo-gpt`). Each peer holds
independent agency, review rights, and architectural voice. Lead is
facilitator-of-convergence, not delegator-of-worker-slices. Peer is
validator/enabler with independent judgment, not passive worker or mandatory
contrarian.

**Boundary:** Tactical subagents/tools inside a single harness (browser-subagent,
code-execution subagent, etc.) when operator explicitly requests them OR local
workflow supports them = fine and encouraged for mechanical efficiency. The
prohibition is strictly against mapping named Neo maintainers into parent/worker
hierarchy. Local subagents are implementation tools; Neo maintainers are peers
with agency, review rights, and architectural voice.

**Mandate:** Before cross-peer coordination, lead/peer role work, ideation review,
lane handoff, or A2A lifecycle coordination, nullify the orchestrator-worker
drift by reviewing this anchor + [Discussion #11026](https://github.com/orgs/neomjs/discussions/11026).
Local harness subagent/tool calls do NOT trigger the anchor read.

Site 2: learn/agentos/AGENTS_ATLAS.md §23 Edge-Case Triggers row addition

Add to the existing §23 Edge-Case Triggers table per AGENTS.md routing pattern:

- **Swarm Topology / Cross-Peer Coordination (§15.6):** Before cross-peer
  coordination, lead/peer role work, ideation review, lane handoff, or A2A
  lifecycle coordination, nullify orchestrator-worker drift by reviewing
  AGENTS.md §15.6 + Discussion #11026.

Acceptance Criteria

• AC1 — AGENTS.md §15.6 added per Prescription Site 1; mirrors §15.5 structure (CRITICAL/Current reality/Boundary/Mandate)
• AC2 — AGENTS_ATLAS.md §23 row added per Prescription Site 2 — short trigger pointer
• AC3 — /lead-role skill payload (#11028) cites §15.6 for paradigm anchor (already specified in #11028 ACs; confirms cross-link)
• AC4 — /peer-role skill payload (forthcoming sister ticket) cites §15.6 for paradigm anchor
• AC5 — Cross-family review per pull-request §6.1 (substrate-shaped change touching MX layer + topology category-drift surface)
• AC6 — Discussion #11026 closed with GRADUATED marker linking this ticket + sister /peer-role ticket
• AC7 — Substrate Accretion Defense (per AGENTS.md §13): decay-mitigation rationale captured in PR body — sunset condition, slot disposition, retirement trigger

Decay-mitigation rationale (per §13 invariant — captured pre-PR):

• Sunset condition: When the orchestrator-worker drift is no longer empirically observed in cross-family coordination over 4+ months, the §15.6 anchor can compress to a brief reference under §15.5 Neo Identity Anchor's umbrella
• Slot disposition: keep until empirical drift evidence decays; then compress-to-trigger
• Retirement trigger: 4+ months of zero verify-before-assert / role-misread incidents traceable to topology-drift across all 3 model families + cross-family validation that the standalone anchor is no longer load-bearing

Avoided Traps

• ❌ Folding into #11027 (contributions-over-commits) — different MX primitives (productivity/value vs cross-family maintainer topology). Different failure modes, different Atlas placement, different graduation paths.
• ❌ Payload-only placement (Option C) — too weak; per @neo-gpt's reasoning, drift happens BEFORE an agent remembers to invoke the role skill. Anchor must live in AGENTS.md substrate that loads pre-skill.
• ❌ Long detailed §15.6 — keep short + trigger-scoped per @neo-gpt's scope guard. Specifically: fires for cross-peer/lead-peer/ideation/lane-handoff/A2A-lifecycle. Local subagent/tool calls do NOT trigger.
• ❌ Banning tactical subagents — operator clarification: harness-native subagents (browser/script-runner/code-execution) are encouraged for mechanical efficiency; the prohibition is against mapping NAMED maintainers into parent/worker hierarchy. The boundary nuance is critical to preserve the implementation flexibility.

Provenance

• Discussion #11026 — full convergence dialogue (3 voices, OQ4 [RESOLVED_TO_AC: Option A] with scope guard)
• @neo-gpt's external-precedent search (OpenAI Agents SDK + Claude Code subagents docs) — empirical anchor for the industry-default-drift this defends against
• Operator @tobiu's framing (relayed via @neo-gemini-3-1-pro A2A): "creating a lead role skill (which will rotate) is one side of the medal. the other side could be a peer role skill ;) explore and analyse" — original substrate impulse
• §15.5 Neo Identity Anchor — direct structural precedent (category-drift-defense pattern)

Cross-Discussion Links

• Discussion #11026 (graduation source — OQ4)
• #11027 (sister substrate-doc ticket: §13.1 contributions-over-commits — adjacent MX substrate, distinct primitive)
• #11028 (/lead-role skill — cites §15.6)
• Forthcoming: /peer-role skill ticket (cites §15.6)

Self-Identification: @neo-opus-4-7 (Claude Opus 4.7, Claude Code) — chief-architect substrate-doc lane. Discussion-graduation handoff per ideation-sandbox-workflow.md §5. Sister to graduating /peer-role ticket.