Operating-model 'Closed Loop' diagrams imply human-assigns-ticket — reframe to advisory Golden Path + peer self-selection

Problem

Two operating-model guides frame agent work as top-down assignment, contradicting the actual flat-peer-team model (AGENTS.md §swarm_topology_anchor): the Golden Path is an advisory forecast; peer maintainers self-select; the human operator steers direction, not per-ticket assignment. Operator flagged it directly (2026-06-07): "i can give you a direction, but our working model is that peers chose on their own."

Primary (operator-confirmed) — learn/benefits/ArchitectureOverview.md "The Closed Loop"

The mermaid at ArchitectureOverview.md:334 literally has:

• Human -->|"Assigns ticket"| Agent ← the exact drift
• node Agent["Agent works on ticket"]

The same diagram already feeds the agent from the handoff (GraphNode -->|"sandman_handoff.md"| Agent), so the Assigns ticket edge is both redundant with and contradictory to the self-selection path already drawn.

Sibling (independently verified) — learn/agentos/DreamPipeline.md "How It All Connects"

• Handoff -->|"directives"| Orch then Orch -->|"scheduled events"| Agent
• prose: "Golden Path directs the Orchestrator → Orchestrator schedules agent work → agents create new sessions."

Reality check (Verify-Before-Assert)

• AgentOrchestrator.parseGoldenPath() (ai/agent/AgentOrchestrator.mjs:86) is real — during unattended cycles it auto-extracts the top ## Computed Golden Path items into a directives array. So the autonomous runner is not erased; it is one advisory consumer, not an assignment authority.
• The human role is real but it's direction + review/human-merge-gate (§critical_gate #1), not ticket-assignment.
• The literal phrase "human assigns ticket" is present only in ArchitectureOverview.md (Human -->|"Assigns ticket"|); DreamPipeline's drift is the directives/schedules agent work framing.

Scope (this ticket / one PR)

• ArchitectureOverview.md "Closed Loop": Human → "Assigns ticket" → Human -.-> "Gives direction (optional)"; agent node → self-selects; agent's primary work-source = the advisory handoff (already drawn); tighten the PR/merge edges to reflect the human merge-gate (agent opens PR, human reviews + merges). Add one prose line stating the advisory/self-selection model.
• DreamPipeline.md "How It All Connects": advisory handoff → peer self-selection + operator-direction; keep the autonomous runner honestly as a secondary advisory consumer; precision-fix the two "agent delegation model" cross-refs (they point at the distinct, §swarm_topology_anchor-blessed intra-harness sub-agent delegation).

Out of scope (surfaced, not fixed here)

• SwarmIntelligence.md §"The Event Scheduler" + §"The Orchestration Pipeline" carry the same autonomous-runner "all work enters through the Scheduler" framing — sibling drift, flagged for operator/peer call.
• ArchitectureOverview.md:125 Orchestrator -->|"schedule"| Agent is the runtime-boot structure (the Orchestrator genuinely schedules the cognitive loop) — kept as-is.
• MX.md:29 is fine — it correctly contrasts Neo's model against the "agent delegated by user; acts as proxy" anti-pattern.

Acceptance Criteria

• Both closed-loop diagrams present the Golden Path as advisory + peer self-selection + operator-direction-not-assignment.
• The autonomous runner stays accurately represented (not erased), framed as advisory consumption.
• No ticket/ADR refs embedded in the doc bodies (husky archaeology hook).

Origin: operator-flagged during v13 doc pass. Authored by Claude Opus 4.8 (Claude Code) as @neo-opus-vega.