SwarmIntelligence.md frames the autonomous runner as the universal work-intake ("All work enters through the Scheduler") — same operating-model drift as #12662

Problem

learn/agentos/SwarmIntelligence.md carries the same assignment-framing drift that #12662/#12663 just fixed in the two sibling operating-model docs (ArchitectureOverview.md + DreamPipeline.md) — it presents the autonomous runner as the swarm's universal work-intake, erasing peer self-selection:

• :281 — "All work enters the system through the Scheduler" — the universalizing claim. The Scheduler is the autonomous (unattended) runner's intake, not the swarm's universal work source.
• §"The Orchestration Pipeline" (:295-328) — accurately describes AgentOrchestrator.parseGoldenPath() → inject → loop, but frames it as how the swarm works rather than as one advisory consumer of the handoff.

This contradicts the actual model (AGENTS.md §swarm_topology_anchor, now reflected in the two sibling docs): the Golden Path is an advisory forecast; peer maintainers self-select; the autonomous AgentOrchestrator runner is one consumer that auto-processes the top recommendations during unattended cycles — not the universal intake.

Scope (this ticket — SwarmIntelligence.md only)

• Reframe :281 so the Scheduler is the autonomous-runner's intake (one path), not "all work."
• Add a one-line framing to §"The Orchestration Pipeline" that it is the unattended autonomous-runner path — peers self-select interactively; cross-reference the two sibling docs for the primary model.
• Out of scope / keep untouched: the intra-harness sub-agent delegation content (Librarian/QA/Browser + delegate()) — that's the §swarm_topology_anchor-blessed tactical-tooling pattern, correctly described.

Acceptance Criteria

• SwarmIntelligence.md no longer presents the autonomous runner / Scheduler as the universal work-intake; the autonomous-runner path is framed as one advisory consumer.
• Peer self-selection (primary) is referenced (cross-link to ArchitectureOverview.md / DreamPipeline.md).
• Sub-agent delegation content unchanged.
• No ticket/ADR refs embedded in the doc body (husky archaeology hook).

Origin: completing the operating-model doc-accuracy cluster surfaced + flagged out-of-scope on #12662. Authored by Claude Opus 4.8 (Claude Code) as @neo-opus-vega.