Problem Statement

The OpenClaw dashboard on port 18789 is the primary user-facing surface after a successful onboard, but no active E2E test verifies that it actually returns HTTP 200. Existing coverage stops at CLI plumbing (nemoclaw list/status/connect) and has no open an HTTP connection to the gateway.
This means regressions in the openclaw gateway process, the OpenShell host → pod port-forward, Landlock hardening, or config-integrity checks can ship to main silently and break the documented quickstart (http://127.0.0.1:18789) without any CI signal.

Proposed Design

Add a E2E script at test/e2e/test-dashboard-reachability.sh that verifies the full host → pod serving chain in one flow.

Flow:

1. Pre-cleanup + prerequisites (Docker, API key, env vars).
2. Run install.sh --non-interactive and assert the sandbox is Ready.
3. Establish the port-forward via openshell forward start --background <port> <sandbox>.
4. Poll http://127.0.0.1:${DASHBOARD_PORT}/ (30 × 1 s) and assert HTTP 200.
5. Body-signature check (HTML grep → <title> → /api/version fallback) to reject unrelated services binding the port.

Alternatives Considered

No response

Category

enhancement: feature

Checklist

• I searched existing issues and this is not a duplicate
• This is a design proposal, not a "please build this" request