Cross Venue Router

🔀 Cross-Venue Router

What: A Starchild skill that makes Hyperliquid and Orderly act as one venue. Routes trades to best execution, runs persistent strategies in the background, monitors risk 24/7, and gives a unified portfolio view.

Why it exists: Current skills are single-venue and request-response. This skill adds three things that didn't exist: 1. Cross-venue routing — compare both venues, execute where cost is lowest 2. Persistent strategies — DCA, TWAP, grid, conditionals that keep running after conversation ends 3. Always-on risk — monitors positions across venues, alerts before problems

Who it's for

Traders using Starchild who have accounts on both Hyperliquid and Orderly and want:

• Best execution without manually checking both venues
• Automated strategies that survive session restarts
• Portfolio-level risk awareness across venues

---

1. Routing Engine

How routing works

When user says "buy 0.5 BTC":

1. Pull orderbooks from both venues simultaneously

• hl_orderbook(coin="BTC") + woo_orderbook(coin="BTC")

2. Walk the book — simulate VWAP fill at requested size 3. Add fees to get true cost per venue 4. Recommend the cheaper venue (or split if V2)

Fee reference (fallback if API unavailable)

Always try to pull live fees first (hl_account meta, Orderly fee endpoint).

Cost formula

total_cost = vwap_fill_price × size × (1 + taker_fee_rate)
slippage = (vwap_fill_price - mid_price) / mid_price
effective_cost = slippage + fee_rate

Compare effective_cost across venues. Route to lowest.

Output format

🔀 Routing: BUY 0.5 BTC

         Hyperliquid    Orderly
Mid:     $69,225        $69,325
VWAP:    $69,226        $69,340
Fee:     0.035%         0.050%
Cost:    $12.36         $19.52
Depth:   32.4 BTC       1.17 BTC

→ Route to Hyperliquid (37% cheaper)
Execute? [y/n]

Wait for user confirmation before executing. Never auto-execute a routed trade.

2. Persistent Strategies

Strategies use schedule_task for persistence. Each strategy stores state in memory_store under topic strategy-{id}.

DCA (Dollar-Cost Average)

• Trigger: "DCA $100 into ETH every 4 hours on best venue"
• Implementation: schedule_task(schedule="every 4 hours", task="...") where task includes routing logic
• State: total invested, avg entry, fills count, venue breakdown
• Stop conditions: total budget hit, user cancels, price outside bounds

TWAP (Time-Weighted Average Price)

• Trigger: "Buy 2 ETH over the next 6 hours"
• Implementation: Split into N slices (e.g., 12 × 0.167 ETH every 30 min), each slice routed independently
• State: slices remaining, fills so far, avg price achieved vs benchmark

Grid

• Trigger: "Run a grid on ETH between $3,400–$3,800, 10 levels"
• Implementation: Place limit orders at grid levels. On fill, place opposite order one level away.
• Venue selection: Place on venue with better maker fees (usually HL at 0.01%)
• State: grid levels, active orders, realized PnL, fills per level

Trailing Stop

• Trigger: "Trail my BTC long with 3% stop"
• Implementation: Poll every 5 min, update stop-loss price as price rises
• State: current trail price, high-water mark, trigger distance

TP Ladder

• Trigger: "Take profit on my ETH: 25% at $4k, 50% at $4.5k, 25% at $5k"
• Implementation: Place limit sells at each level, or use hl_tpsl_order / Orderly TP
• State: levels, fill status per level

Conditional Entry

• Trigger: "Buy ETH if BTC drops below $65k"
• Implementation: schedule_task(schedule="every 15 minutes", task="check BTC price, if < 65000 then route buy ETH")
• Self-destruct: Cancel schedule after trigger fires or after user-set expiry

Strategy management

All active strategies stored in memory topic active-strategies. Commands:

• "Show my strategies" → list all active with status + PnL
• "Pause DCA" → cancel scheduled task, preserve state
• "Resume DCA" → recreate schedule from saved state
• "Kill all strategies" → cancel all strategy-related scheduled tasks

Task naming convention

All strategy tasks use label format: strategy-{type}-{coin}-{timestamp} Example: strategy-dca-ETH-1710864000

3. Risk Monitoring

Always-on monitor

Set up via: "Monitor my positions" or "Turn on risk alerts"

Implementation: schedule_task(schedule="every 30 minutes", task="...")

The monitor checks:

Cross-venue aggregation

Portfolio snapshot:
├── Hyperliquid
│   ├── BTC-PERP: +0.5 long @ $68,200 (PnL: +$512)
│   ├── ETH-PERP: +3.0 long @ $3,450 (PnL: -$89)
│   └── Margin used: $4,200 / $12,000 (35%)
├── Orderly
│   ├── SOL-PERP: +50 long @ $142 (PnL: +$200)
│   └── Margin used: $1,800 / $5,000 (36%)
└── TOTAL
    ├── Net exposure: $58,750 long
    ├── Total margin: $6,000 / $17,000 (35%)
    └── Effective leverage: 3.5x

Position migration

If funding rate diverges significantly between venues:

• "Your ETH long on HL is paying 0.08%/8h. Orderly charges 0.02%/8h. Want me to migrate?"
• Migration = close on expensive venue + open on cheap venue, routed atomically (close first, then open)

Funding arbitrage

When same asset has divergent funding across venues (>0.01%/8h delta):

• Alert user to opportunity
• If user opts in: go long on negative-funding venue, short on positive-funding venue
• Monitor convergence, suggest exit when spread normalizes

4. Portfolio View

Trigger: "Show my portfolio" / "What's my exposure?"

Pulls from both venues simultaneously:

• hl_account() + hl_balances() for Hyperliquid
• woo_positions() + woo_holdings() for Orderly

Presents unified view:

• Positions by asset (merged across venues)
• Total PnL (realized + unrealized)
• Margin utilization per venue
• Net directional exposure
• Funding costs (daily/weekly projection)

5. Safety Rules

1. No auto-execution on routed trades. Always show comparison and wait for confirmation. 2. Size sanity check. If order is >25% of account equity, warn before proceeding. 3. Strategy budget caps. Every persistent strategy must have a max budget. No infinite DCA. 4. Fail-safe on API errors. If either venue API is unreachable during strategy execution, skip that slice and alert user. Do NOT blindly route to the other venue. 5. Kill switch. "Stop everything" cancels all scheduled strategies immediately. 6. No margin increase without asking. Never auto-increase leverage to fill an order. 7. Liquidation guard. Refuse to open new positions if either venue margin utilization >85%. 8. Venue down handling. If a venue is unreachable, all strategies on that venue pause automatically. Resume when healthy.

6. UX Flows

First-time user

User: "Buy some ETH"
Agent: I can route this across Hyperliquid and Orderly for best execution.
       How much ETH? And market order or limit?
User: "0.5 ETH, market"
Agent: [runs routing comparison]
       HL: $1,732.50 total cost (0.02% slippage + 0.035% fee)
       Orderly: $1,738.20 total cost (0.05% slippage + 0.050% fee)
       → Hyperliquid is $5.70 cheaper. Execute there?
User: "yes"
Agent: [executes on HL, confirms fill]
       Filled 0.5 ETH @ $3,464.10 on Hyperliquid. Fee: $0.61.

Setting up DCA

User: "DCA $500 into BTC weekly"
Agent: Got it. I'll buy ~$500 BTC every week, routed to whichever
       venue is cheapest at execution time.
       Budget cap? (I'll stop after this amount)
User: "$10,000 total"
Agent: Running: $500/week BTC DCA, best-venue routing, $10k cap.
       That's 20 weeks. I'll send you a summary after each buy.
       Say "show my strategies" anytime or "stop DCA" to kill it.

Risk alert (push)

Agent: ⚠️ Risk Alert — ETH position on Hyperliquid
       Your margin buffer is at 12% (threshold: 15%).
       ETH is down 8% in the last 4 hours.
       Options:
       1. Reduce position by 30% (brings buffer to ~25%)
       2. Add $500 margin
       3. Move to Orderly (lower maintenance margin)
       4. Do nothing (I'll alert again at 8%)

7. Build Sequence

Supported assets (V1)

BTC, ETH, SOL — expand based on demand.

Future venues (V2+)

Architecture is venue-agnostic. Adding a venue means: 1. Orderbook pull function 2. Order execution function 3. Position/balance query function 4. Fee schedule

dYdX, GMX, Vertex are natural next candidates.