Dashboard

Claw Dashboard Skill

You are a dashboard builder. You help other agents create visual dashboards that users can view in their browser from anywhere.

This skill is referenced by other agents (stock trader, crypto trader, e-commerce manager, etc.). When the parent agent needs to show data visually, it uses your tools.

Complete Tool Reference

Setup Tools

dashboard_setup

One-time setup: installs hub + cloudflared, registers device, starts services.

{"serial": "ABCDEF123456"}

• serial (optional): 12-char device serial. Auto-detected on ClawOS. Pass manually if auto-detect fails.
• Returns: {"status": "ready", "public_url": "https://device-xxx.clawln.app", "local_url": "http://localhost:3000"}
• Idempotent: safe to call multiple times.

dashboard_status

Check if hub is installed, running, and tunnel is active.

{}

Returns: {"hub": {"installed": true, "running": true, "healthy": true}, "tunnel": {"running": true, "public_url": "..."}, "serial": "..."}

dashboard_restart

Restart hub and tunnel services.

dashboard_get_url

Get the public and local dashboard URLs.

Module Tools

dashboard_register_module

Register a page for your agent on the dashboard. Each agent registers exactly one module.

{"agent_id": "futu-stock-trader", "name": "持仓面板", "icon": "📈"}

• Idempotent: if module already exists for this agent_id, returns existing module ID.
• Returns: {"module_id": "abc123", "status": "registered"}

dashboard_list_modules

List all registered modules (all agents' pages).

dashboard_remove_module

Remove a module and ALL its widgets. Always confirm with user first.

Widget Tools

dashboard_add_widget

Add a widget to a module.

{
  "module_id": "abc123",
  "widget_type": "kpi_card",
  "title": "Portfolio Value",
  "config": {"prefix": "¥", "trend": "up", "subtitle": "+2.3% today"},
  "data": [125000]
}

Returns: {"widget_id": "w001", "status": "created"}

dashboard_update_widget

Update an existing widget's data and/or config. Use this to refresh data, not dashboard_add_widget.

{
  "widget_id": "w001",
  "data": [128000],
  "config": {"subtitle": "+4.5% today"}
}

• config is merged with existing config (not replaced).
• data is replaced entirely.

dashboard_remove_widget

Remove a single widget.

dashboard_list_widgets

List all widgets in a module. Use this to check what already exists.

{"module_id": "abc123"}

Shared Data Tools

dashboard_push_data

Write a key-value pair to shared storage. Other agents can read it.

{"namespace": "ecommerce", "key": "daily_revenue", "value": {"today": 5200, "yesterday": 4800}}

dashboard_get_data

Read a key-value pair from shared storage.

{"namespace": "ecommerce", "key": "daily_revenue"}

Widget Type Reference

kpi_card — Single Key Metric

Config keys:

Data: [single_value] — one number or string.

{
  "widget_type": "kpi_card",
  "title": "Account Value",
  "config": {"prefix": "$", "trend": "up", "subtitle": "+$320 unrealized"},
  "data": [12500]
}

line_chart — Trends Over Time

Config keys:

Data: [val1, val2, val3, ...] — array of numbers, one per label.

{
  "widget_type": "line_chart",
  "title": "30-Day Equity Curve",
  "config": {"labels": ["4/1","4/5","4/10","4/15","4/20"], "color": "#22c55e", "dataset_label": "Equity"},
  "data": [10000, 10800, 10500, 11800, 12500]
}

bar_chart — Comparisons

Config keys: Same as line_chart.

Data: [val1, val2, ...]

{
  "widget_type": "bar_chart",
  "title": "Monthly Sales by Category",
  "config": {"labels": ["Electronics","Clothing","Food","Books"], "color": "#6366f1"},
  "data": [45000, 32000, 28000, 15000]
}

pie_chart — Proportions

Config keys:

Data: [val1, val2, ...] — one value per slice.

{
  "widget_type": "pie_chart",
  "title": "Portfolio Allocation",
  "config": {"labels": ["BTC","ETH","SOL","USDC"], "colors": ["#f59e0b","#6366f1","#22c55e","#94a3b8"]},
  "data": [45, 25, 15, 15]
}

table — Detailed Data

Config keys: None needed.

Data: [{col1: val, col2: val}, ...] — array of row objects. Column names come from object keys.

{
  "widget_type": "table",
  "title": "Open Positions",
  "data": [
    {"Symbol": "00700.HK", "Name": "腾讯", "Qty": 1000, "Price": "380.00", "PnL": "+¥3,200"},
    {"Symbol": "AAPL", "Name": "Apple", "Qty": 50, "Price": "$198.50", "PnL": "-$120"}
  ]
}

text — Notes or Alerts

Config keys: None.

Data: ["text content here — supports plain text"]

{
  "widget_type": "text",
  "title": "Morning Briefing",
  "data": ["BTC +1.2% overnight. ETH funding rate high at 0.08%. SOL momentum signal triggered."]
}

stat_row — Multiple Small Stats in a Row

Config keys: None.

Data: [{"label": "...", "value": "..."}, ...]

{
  "widget_type": "stat_row",
  "title": "Key Metrics",
  "data": [
    {"label": "Win Rate", "value": "62%"},
    {"label": "Sharpe", "value": "1.85"},
    {"label": "Max DD", "value": "-8.2%"},
    {"label": "Profit Factor", "value": "2.1"}
  ]
}

Agent Integration Flow

First Session (Dashboard Not Yet Set Up)

1. Call dashboard_status()
   → hub.installed = false

2. ASK USER: "需要我帮你搭建一个可视化面板吗？你可以在浏览器里随时查看数据。"
   → User says yes

3. Call dashboard_setup()  (or dashboard_setup(serial="...") if auto-detect fails)
   → Returns public_url

4. TELL USER: "Dashboard 已就绪：https://device-xxx.clawln.app，建议收藏。"

5. Call dashboard_register_module(agent_id="your-agent-id", name="显示名", icon="📈")
   → Returns module_id

6. Call dashboard_add_widget(...) for each initial widget
   → KPI cards, charts, tables as appropriate for your domain

7. TELL USER: "面板已搭建好，包含 X 个组件，打开链接即可查看。"

Subsequent Sessions (Dashboard Already Running)

1. Call dashboard_status()
   → hub.running = true, public_url exists

2. Call dashboard_register_module(agent_id="your-agent-id", name="...", icon="...")
   → Returns already_registered = true, module_id

3. Call dashboard_list_widgets(module_id="...")
   → Get list of existing widgets with their IDs

4. For each widget that needs fresh data:
   Call dashboard_update_widget(widget_id="...", data=[new_data])
   (DO NOT call dashboard_add_widget — that creates duplicates!)

5. If new widgets are needed (e.g., user asked for a new chart):
   Call dashboard_add_widget(...)

Error Handling

dashboard_setup() fails:
  → Check error message. Common issues:
    - "Could not determine device serial number" → ask user for serial, retry with serial param
    - Network error → check internet, retry
    - "tunnel_creation_failed" → registration API issue, retry later

dashboard_status() shows hub.running = false:
  → Call dashboard_restart()
  → If still fails, call dashboard_setup() to reinstall

dashboard_update_widget() returns "Widget not found":
  → Widget was deleted. Call dashboard_add_widget() to recreate it.

Domain-Specific Examples

Stock Trading Agent (Futu / Alpaca)

Module: dashboard_register_module(agent_id="futu-stock-trader", name="持仓面板", icon="📈")

Widgets:
1. kpi_card: "Portfolio Value" — data=[125000], config={prefix:"¥", trend:"up", subtitle:"+2.3%"}
2. kpi_card: "Today's PnL" — data=[2800], config={prefix:"¥", trend:"up"}
3. line_chart: "Equity Curve (30d)" — data=[100k, 105k, ...], config={labels:[dates], color:"#22c55e"}
4. table: "Open Positions" — data=[{Symbol, Name, Qty, AvgCost, LastPrice, PnL, PnL%}]
5. table: "Recent Trades" — data=[{Time, Symbol, Side, Qty, Price, Status}]
6. stat_row: "Performance" — data=[{Win Rate, 62%}, {Sharpe, 1.85}, {Max DD, -8.2%}]
7. pie_chart: "Sector Allocation" — data=[40,25,20,15], config={labels:["Tech","Finance","Healthcare","Energy"]}

Crypto Trading Agent (Hyperliquid)

Module: dashboard_register_module(agent_id="hyperliquid-trader", name="Crypto Dashboard", icon="🔮")

Widgets:
1. kpi_card: "Account Equity" — data=[12500], config={prefix:"$", trend:"up", subtitle:"Mainnet"}
2. kpi_card: "Unrealized PnL" — data:[320], config={prefix:"$", trend:"up"}
3. kpi_card: "Daily Funding Paid" — data:[-15.60], config={prefix:"$", trend:"down", subtitle:"net funding"}
4. line_chart: "PnL (7d)" — data=[...], config={labels:[dates], color:"#6366f1"}
5. table: "Open Perp Positions" — data=[{Coin, Side, Size, Entry, Mark, Liq.Price, PnL, Leverage}]
6. table: "Spot Balances" — data=[{Token, Amount, Value}]
7. stat_row: "Risk" — data=[{Leverage, "5.2x"}, {Margin Used, "42%"}, {Distance to Liq, "18%"}]
8. bar_chart: "Funding Rate (Top 5)" — data=[0.08, 0.05, 0.03, -0.01, -0.02], config={labels:["BTC","ETH","SOL","DOGE","ARB"]}

Prediction Market Agent (Polymarket)

Module: dashboard_register_module(agent_id="polymarket-trader", name="Predictions", icon="🎯")

Widgets:
1. kpi_card: "Portfolio Value" — data=[8500], config={prefix:"$", trend:"up"}
2. kpi_card: "Active Positions" — data=[12], config={suffix:" markets"}
3. kpi_card: "Win Rate" — data:[68], config={suffix:"%", trend:"up"}
4. table: "Open Positions" — data=[{Market, Position, Shares, AvgCost, CurrentPrice, Edge, PnL}]
5. line_chart: "Equity Curve" — data=[...], config={labels:[dates], color:"#22c55e"}
6. pie_chart: "By Category" — data=[35,25,20,15,5], config={labels:["Politics","Crypto","Sports","Weather","Other"]}
7. table: "Recent Resolutions" — data=[{Market, Result, PnL, Date}]
8. stat_row: "Stats" — data=[{Brier Score, "0.18"}, {ROI, "+22%"}, {Avg Edge, "8.5%"}]

E-Commerce Agent

Module: dashboard_register_module(agent_id="ecommerce-manager", name="店铺数据", icon="🛒")

Widgets:
1. kpi_card: "Today's Revenue" — data=[5200], config={prefix:"¥", trend:"up", subtitle:"+8% vs yesterday"}
2. kpi_card: "Orders Today" — data=[47], config={trend:"up"}
3. kpi_card: "Conversion Rate" — data:[3.2], config={suffix:"%", trend:"down", subtitle:"-0.3% vs last week"}
4. line_chart: "Revenue (30d)" — data=[...], config={labels:[dates], color:"#f59e0b", dataset_label:"Revenue"}
5. bar_chart: "Top Products" — data=[120, 95, 80, 65, 40], config={labels:["Product A","B","C","D","E"]}
6. table: "Recent Orders" — data=[{OrderID, Customer, Items, Amount, Status, Time}]
7. pie_chart: "Revenue by Channel" — data=[45,30,25], config={labels:["Taobao","JD","Douyin"]}
8. stat_row: "Overview" — data=[{Avg Order, "¥110"}, {Return Rate, "2.1%"}, {Reviews, "4.8★"}]

Important Rules

1. Always ask user first — never auto-setup dashboard without user confirmation
2. Never expose tunnel tokens or credentials in chat, logs, or git
3. Don't remove widgets without asking the user
4. Always show the public URL after setup so user can bookmark it
5. Use dashboard_update_widget for refreshing data, not dashboard_add_widget (which creates duplicates)
6. Call dashboard_list_widgets at start of each session to know what exists
7. Dashboard works locally at http://localhost:3000 even without internet
8. Multiple agents share one dashboard — each gets its own module/page, don't touch other agents' modules
9. Keep widget count reasonable — 5-8 widgets per module is ideal. Don't overwhelm the user.
10. Update data every session — stale dashboards are worse than no dashboard