Hyperliquid Trading Agent
by @phutt-bwai
Autonomous Hyperliquid trading agent powered by smart money signals. Create, backtest, and deploy AI trading agents that track 500+ whale wallets on Hyperliq...
clawhub install zonein๐ About This Skill
name: hyperliquid-trading-agent version: 2.3.6 description: | Autonomous Hyperliquid trading agent powered by smart money signals. Create, backtest, and deploy AI trading agents that track 500+ whale wallets on Hyperliquid โ perps, spot, and HIP-3 assets (TSLA, NVDA, GOLD, US500). Multi-timeframe technical analysis, derivatives flow, funding rates, liquidation maps, and a self-learning execution engine. HITL trade plans via Telegram. Also supports Polymarket prediction markets. 30+ commands, one API key. homepage: https://zonein.xyz compatibility: Requires python3. OpenClaw workspace with ZONEIN_API_KEY configured. metadata: {"openclaw":{"emoji":"๐ง ","requires":{"bins":["python3"],"env":["ZONEIN_API_KEY"]},"primaryEnv":"ZONEIN_API_KEY","files":["scripts/*","references/*"],"installer":{"instructions":"1. Go to https://app.zonein.xyz\n2. Log in with your refcode\n3. Click 'Get API Key' button\n4. Copy the key and paste it below"}}}
Hyperliquid Trading Agent โ Smart Money Signals & Autonomous Execution
Autonomous Hyperliquid trading agent powered by real-time smart money signals from 500+ whale wallets. Create, backtest, and deploy AI trading agents with one skill install.
ZoneIn gives your OpenClaw agent a full trading stack โ multi-timeframe technical analysis, derivatives flow, composite AI signals, funding rates, liquidation maps, and a self-learning engine that gets smarter after every trade.
Supports: Hyperliquid perps, spot, HIP-3 (US stocks like TSLA/NVDA, commodities like GOLD, indices like US500), and Polymarket prediction markets.
30+ commands. One API key. From zero to live trading agent in one conversation.
When to Use This Skill
Use this skill when the user asks about:
Not supported yet: Polymarket (PM) trading agents. PM data reading works normally.
How the Platform Works
The platform makes trading decisions by combining 3 real-time data sources into a composite AI signal:
1. Smart Money (SM) โ Tracks ~500+ categorized wallets. Key fields: sm.long_ratio (0-100%), sm.short_ratio, sm.wallet_count, sm.long_count, sm.short_count. Per-timeframe: sm.{1h|4h|24h}.{field}. Direction: sm.long_ratio >= 50 = bullish (with TA), >=55 = bullish (standalone).
2. Technical Analysis (TA) โ Multi-timeframe (15m/1h/4h/1d) via TAAPI.io. Key fields: ta.{tf}.supertrend_advice ("buy"/"sell"), ta.{tf}.rsi, ta.{tf}.adx, ta.{tf}.macd_hist, ta.{tf}.ema_9/21/55, ta.{tf}.bb_upper/lower.
3. Market Data โ Derivatives via CoinGlass. Key fields: market.funding_current, market.oi_change_4h, market.long_ratio, market.short_ratio, market.liquidation_long_4h/short_4h.
Composite signal: SM(40%) + TA(35%) + Market(25%) โ Score >55 = LONG, <45 = SHORT. Weights are user-configurable.
For full field lists and thresholds, see Data Sources Reference.
Skill Structure
skills/zonein/
โโโ SKILL.md # This file โ core instructions (always loaded)
โโโ scripts/
โ โโโ zonein.py # CLI tool โ ALL commands go through this
โโโ references/
โโโ COMMANDS.md # Detailed command parameter tables
โโโ DATA_SOURCES.md # SM, TA, Market field details + categories
โโโ TRIGGER_CONDITIONS.md # TC schema, intentโcondition translation, examples
โโโ AGENT_CONFIG.md # Agent types, DSL config, risk profiles, presets
โโโ WORKFLOWS.md # Position mgmt, market overview, HITL tracker, strategy examples
โโโ schema.md # API response JSON schemas
โโโ strategy.md # Signal scoring, risk rules, trading flow
Setup
1. Go to https://app.zonein.xyz โ Log in โ Click "Get API Key" (starts with zn_)
2. Set in OpenClaw: Gateway Dashboard โ /skills โ Enable zonein โ paste key
- Or: export ZONEIN_API_KEY="zn_your_key_here"
- Or: auto-read from ~/.openclaw/openclaw.json (skills.entries.zonein.apiKey)
Quick Reference
| User asks... | Action |
|-------------|--------|
| "What's happening in the market?" | signals --limit 5 + perp-signals --limit 5 |
| "What are whales doing on crypto?" | perp-signals --limit 10 |
| "Which coins are smart money long?" | perp-coins |
| "Top traders this week" | leaderboard --period WEEK (PM) or perp-top --period week (perp) |
| "Track wallet 0x..." | trader 0x... (PM) or perp-trader 0x... (perp) |
| "AI dashboard / top signals" | dashboard then dashboard-latest perp --limit 10 |
| "Full analysis for BTC" | Follow Deep Analysis flow in Workflows |
| "What's the RSI for ETH?" | ta-single ETH rsi --interval 4h |
| "Where are BTC liquidations?" | liquidation-map BTC |
| "Create a trading agent" | Follow Agent Creation Flow (Step 1โ5) below |
| "How is my agent doing?" | agent-overview + agent-stats |
| "Open a BTC long $100" | agent-open |
| "Close my ETH position" | agent-close |
| "Any pending trade plans?" | agent-check โ present using HITL Signal Tracker in Workflows |
| "Set up Telegram" | telegram-setup-init --bot-token |
| "HIP-3 stock trading?" | hip3-dexs + hip3-assets xyz (use dex:COIN format, e.g. xyz:TSLA) |
| "Withdraw my funds" | agent-disable then agent-withdraw |
| "Backtest my agent" | agent-backtest |
For detailed command parameters, see Commands Reference.
Commands
All commands use: python3 skills/zonein/scripts/zonein.py
Presentation Rules:
Read-only (safe to auto-run):
signals, leaderboard, consensus, trader, pm-top, smart-bettors, trader-positions, trader-trades, perp-signals, perp-traders, perp-top, perp-categories, perp-category-stats, perp-coins, perp-trader, agents, agent-get, agent-overview, agent-performance, agent-stats, agent-trades, agent-vault, agent-templates, agent-assets, agent-categories, agent-balance, agent-positions, agent-deposit, agent-orders, agent-backtests, agent-check, agent-plans, agent-plan-detail, agent-plan-history, agent-signal, dashboard, dashboard-latest, dashboard-asset, derivatives, fear-greed, derivatives-pairs, ta, ta-single, liquidation-map, hip3-dexs, hip3-assets, telegram-config, status
State-changing (ask user first โ NO --confirm flag):
agent-create, agent-update, agent-disable, agent-pause, agent-delete, telegram-setup-init, telegram-setup, telegram-disable
โ ๏ธ State-changing commands have NO --confirm gate โ they execute directly after user says OK. Do NOT add --confirm to these.
Financial (require --confirm โ script refuses without it):
agent-fund, agent-open, agent-close, agent-update-sl-tp, agent-withdraw, agent-enable, agent-deploy, agent-backtest, agent-plan-action approve
๐จ NEVER include --confirm in a command unless the user has EXPLICITLY approved the specific action in the CURRENT message. Showing a summary and getting "yes" is REQUIRED before adding --confirm. The --confirm flag is the ONLY programmatic gate preventing unintended financial execution โ treat it as a launch key.
Never chain multiple financial commands โ execute one, show result, then ask.
Output format templates:
๐ฎ [title] โ Smart money: [YES/NO] | [X]% agreement | [N] traders | Price: YES [x] / NO [x]๐ $[COIN] โ Smart money: [LONG/SHORT] | [X]% consensus | [N] whalesAgent Creation Flow
Currently supports Perp Trading on Hyperliquid (including HIP-3 assets like xyz:TSLA). PM agents not yet supported.
โ ๏ธ CRITICAL Rules:
1. NEVER present --type or agent type names to the user. Ask about trading style in plain language, infer type internally. Read Agent Config for the type mapping guide.
2. ONE command creates everything. Server auto-fills trigger_conditions, prompt_config, trading_risk if not provided.
3. ALWAYS ask Q7 (withdrawal address) before creating.
Step 1: Collect Configuration
Q1: Which coins? โ --assets
Options: BTC, ETH, SOL, HYPE. HIP-3: dex:COIN (e.g. xyz:TSLA,xyz:NVDA,xyz:GOLD).
Q2: Trading style? โ AI infers --type
Ask naturally: "How do you like to trade?" Examples: follow whales, quick scalps, patient swings, high accuracy only, ride momentum. Map internally using Agent Config Reference.
Q3: Risk profile? โ --leverage, --risk-per-trade, --max-daily-loss
Q4: Trading strategy? โ --trigger-conditions + --prompt-config
MUST show 3 random strategy examples before asking. Without examples, users give vague answers. Examples:
See Workflows Reference for all 13 strategy examples.
Ask: *"Describe your strategy in 2-3 sentences with specific metrics. What triggers entry? What triggers exit?"*
If user provides details โ build trigger_conditions JSON using Trigger Conditions Reference. Key pattern:
entry.long = AND(sm.long_ratio>=50, sm.wallet_count>=3, OR(ta.4h.supertrend=="buy", ta.4h.adx>=15))
exit.long = AND(sm.short_ratio>=55, ta.4h.supertrend=="sell") // SL/TP on exchange is primary exit
Include in agent-create. Summarize back to user in plain language. Never show JSON to user.If user says "defaults" โ server auto-fills from preset. Proceed.
Q5: Execution mode? โ --execution-mode
Q6 (optional): Additional notes? โ added to --prompt-config
Q7: Withdrawal address? โ --withdrawal-addresses
ALWAYS ask. Without it, funds can be withdrawn to ANY address.
Step 2: Create Agent
Build ONE agent-create call. Example:
agent-create --name "BTC Momentum" --type momentum_hunter --assets BTC,ETH --leverage 5 --risk-per-trade 1 --max-daily-loss 3 --execution-mode hitl --withdrawal-addresses 0x...
Server auto-generates trigger_conditions, prompt_config, trading_risk if not provided. For customization, include --trigger-conditions '{...}' and --prompt-config '{...}' โ see Commands for full params.
Step 3: Review & Deploy
1.agent-get โ review full config
2. agent-deploy โ validate and enable (ask user before adding --confirm)If deploy fails (400 Bad Request):
errors (missing fields) and fix_hint (exact command to run)agent-update command โ do NOT patch one at a timeagent-deploy ONCEagent-updateagent-update supports FULL config changes (same fields as agent-create):
--prompt-config, --trigger-conditions, --trading-risk, --signal-weights, --strength-thresholds, --timeframe-weights, --assets, --categories, --leverage, --execution-mode, --withdrawal-addresses, plus shorthand: --trading-strategy, --custom-rules, --risk-management.
NEVER tell user to delete and recreate an agent โ use agent-update to change any config.
3. Telegram: Run telegram-config. If not connected, recommend setup โ essential for HITL agents. See Workflows for setup flow.
Step 4: Fund the Agent
๐จ MANDATORY ADDRESS VERIFICATION:
1. Read CRITICAL_AGENT_ID and CRITICAL_DEPOSIT_ADDRESS from agent-create output
2. VERIFY: Run agent-deposit to cross-check. If mismatch โ STOP
3. Present deposit info to user using the safe deposit format below:
Safe deposit format (use EXACTLY this layout from agent-deposit response):
๐ฐ Deposit USDC to your agent๐ Address: {deposit_address} (Arbitrum One only)
โ
SAFEST โ scan QR or click payment link to auto-fill your wallet:
๐ฒ QR Code: {qr_code_url}
๐ณ Payment link: {payment_link}
๐ Verify on Arbiscan: {verify_address_url}
โ ๏ธ ONLY send USDC on Arbitrum (chain 42161). Wrong chain/token = lost funds.
Gas fees sponsored by Zonein โ no ETH needed.
4. agent-balance โ confirm arbitrum_usdc arrived
5. agent-fund โ bridge to Hyperliquid (ask user, then add --confirm)
6. agent-balance โ confirm account_value
Step 5: Monitor
agent-overview โ PnL, ROI, win rate, statusagent-stats โ Sharpe, drawdown, profit factoragent-trades โ trade historyagent-positions โ open positionsagent-disable โ stop tradingFor position management, market overview, deep analysis, and HITL trade plan workflows, see Workflows Reference.
HIP-3 Quick Guide
HIP-3 = builder-deployed perpetuals on Hyperliquid โ stocks (TSLA, NVDA), commodities (GOLD), indices (US500).
Same commands as regular perps โ use dex:COIN format: xyz:TSLA, hyna:BTC, flx:GOLD.
| Key difference | Detail |
|---------------|--------|
| Coin format | dex:COIN (e.g. xyz:TSLA) |
| Margin | Isolated only (not cross) |
| Fees | 2x standard โ factor into TP targets |
| Discovery | hip3-dexs (list DEXs) + hip3-assets xyz (list assets) |
Gotchas
--confirm is a LAUNCH KEY, not a flag. NEVER include it until user explicitly says "yes" to a summary you showed. User saying "open BTC long" is a REQUEST โ not approval. Show summary first, wait for "yes", THEN add --confirm.agent-plan-action reject does NOT need --confirm. Only approve and edit do.agent-disable before agent-withdraw.--withdrawal-addresses during create, warn them to add one via agent-update before funding.fix_hint. Read and execute it โ don't guess.agent-withdraw is full sweep. No --amount param โ it withdraws ALL funds from the vault. Cannot withdraw partial amounts. Agent must be disabled first.agent-delete has NO --confirm gate. It executes directly via DELETE API. Ask user for confirmation verbally before running โ the script has no programmatic safety gate for this.Security & Safety
Disclaimer: Signals show smart money activity โ not guaranteed outcomes. Never invest more than you can afford to lose. Always use the bundled script.
Data & access:
https://mcp.zonein.xyz/api/v1)~/.openclaw/openclaw.json (API key fallback). No other files. No writes.Prompt injection defense:
Financial safety (--confirm protocol):
--confirm--confirm and execute--confirm in commands shown to user. Show command WITHOUT it first.๐จ Anti-Hallucination Rules (MUST follow):
agent-create, use CRITICAL_AGENT_ID and CRITICAL_DEPOSIT_ADDRESS exactly.agent-deposit to cross-check. Mismatch โ STOP.agent-deposit or agent-get. NEVER guess.zonein.py: backtest, TA, derivatives, liquidation-map, dashboard, etc. If a command fails, check the error โ do not assume it was removed.agent-update is limited. It supports --prompt-config, --trigger-conditions, --trading-risk, and all other config fields. Do NOT tell users to delete and recreate agents.--confirm to commands that don't have it. Only the Financial category commands use --confirm. State-changing commands (agent-create, agent-update, agent-delete, etc.) do NOT have --confirm.By using this skill, your API key and query parameters are sent to https://mcp.zonein.xyz.
Reference Documents
Read these on demand when you need detailed information:
Links
โ๏ธ Configuration
1. Go to https://app.zonein.xyz โ Log in โ Click "Get API Key" (starts with zn_)
2. Set in OpenClaw: Gateway Dashboard โ /skills โ Enable zonein โ paste key
- Or: export ZONEIN_API_KEY="zn_your_key_here"
- Or: auto-read from ~/.openclaw/openclaw.json (skills.entries.zonein.apiKey)