dashboardlisted
Install: claude install-skill x-mesh/xm
<Purpose>
Start, stop, or check the xm-dashboard web server that visualizes .xm/ project state.
</Purpose>
<Use_When>
- User says "dashboard", "open dashboard", "start dashboard", "show dashboard"
- User says "stop dashboard", "close dashboard", "kill dashboard"
- User says "dashboard status"
</Use_When>
<Do_Not_Use_When>
- User wants to query .xm data directly (use x-build status instead)
</Do_Not_Use_When>
# x-dashboard
## Model Routing
This entire skill is **haiku** (Agent tool). All commands (start/stop/status/open) are pure script execution — bun process management, curl health checks, browser open. Zero reasoning required.
| Command | Model | Reason |
|---------|-------|--------|
| `start` | **haiku** | nohup + sleep + curl |
| `stop` | **haiku** | bun --stop |
| `status` | **haiku** | curl + JSON display |
| `open` | **haiku** | macOS open command |
```
Agent tool: { model: "haiku", description: "x-dashboard <cmd>", prompt: "Run: <bash from command section>" } <!-- managed-model: writer -->
```
**Guardrail**: never haiku if the user asks "why is the dashboard showing X" or "interpret these metrics" — interpretation is sonnet-level reasoning.
## Arguments
User provided: $ARGUMENTS
## Routing
Parse `$ARGUMENTS`:
- `stop` / `close` / `kill` → [Command: stop]
- `restart` / `relaunch` / `reload` → [Command: restart]
- `status` → [Command: status]
- `open` → [Command: open]
- Empty or `start` or any other text → [Command: start]
## Command: start
1. Check if a