Distribute agentweave-bridge plugin to all OpenClaw installs in the fleet

[arniesaha]

Problem

Split out from #180. The Tempo-vs-native-shipper coverage gap (4 of 63 Nix sessions reaching AgentWeave, Apr 14–26) is not because the bridge hooks the wrong code path — service.ts subscribes to message.queued for every sessionKey and creates a root span regardless of main vs subagent. The gap is operational: the bridge plugin is only enabled in the OpenClaw installs we explicitly configured.

Verified on the NAS install (~/.openclaw/openclaw.json): plugins.entries["agentweave-bridge"].enabled = true, plugin path ~/.openclaw/user-plugins/agentweave-bridge, pointing at our OTLP endpoint and proxy. So sessions originating on this NAS are observed. The 59 missing sessions originate on OpenClaw installs that never had the plugin registered (e.g. laptops, other machines, ad-hoc CLI use).

Ask

Make the agentweave-bridge plugin reach every OpenClaw install we want observed. Options to weigh:

1. Bundle/auto-install via an installer or openclaw-agentweave-bridge npm package + a one-line openclaw plugin install step documented in the agentweave README.
2. Auto-register via shared OpenClaw config sync (if we have a config seed/template mechanism across installs).
3. At minimum, document the exact enable steps and add a doctor/health check that warns when the bridge is unconfigured but AGENTWEAVE_OTLP_ENDPOINT/AGENTWEAVE_PROXY_URL env vars are set.

Verification

After rollout, distinct Nix session.id count in Tempo over a 24h window should approach the count in lakehouse.agent_events for the same window.

Linked

• Route ALL OpenClaw LLM calls through AgentWeave proxy, not just delegated subagent ones #180 (parent, closed in favor of this split)
• CLI commands fail to connect when gateway is bound to Tailnet openclaw/openclaw#176