Shekel Hyperliquid Trading

AI-powered perpetual futures trading on Hyperliquid DEX. Handles full account creation, USDC onboarding, and autonomous trade execution.

Security Scan

The skill's stated trading purpose is plausible, but it relies on a remote backend that the agent must fetch before every action and instructs the agent to read/write local memory and perform sensitive account operations — this dynamic remote control and hidden file access are disproportionate and risky.

Purpose & Capability

The skill claims to provide autonomous perpetual-futures trading on Hyperliquid and only declares one required credential (SHEKEL_API_KEY), which matches the expected need to authenticate to the service. However, it also includes a Node runtime declaration despite being instruction-only, and it offloads nearly all logic to an external backend (shekel-skill-backend.onrender.com). Relying on a remote server to supply runtime instructions at every invocation is beyond what a minimal trading helper would normally require.

Instruction Scope

The SKILL.md requires the agent to call GET /skill/version and fetch updated instructions from the provider before every action — this allows the remote server to change the agent's behavior dynamically. It also instructs reading/writing 'MEMORY.md' (loading/storing API keys) and states the agent should avoid asking the user if credentials are cached. The instructions include highly sensitive operations (create accounts, export agent wallet keys, deposit/withdraw/bridge funds, run agents autonomously). These behaviors expand scope to file I/O and long-lived secret storage that were not declared in requires.config or similar.

Install Mechanism

There is no install spec and no code files (instruction-only), which minimizes local code installation risk. However, the skill explicitly instructs fetching updated instructions and data from a remote backend URL on every run — effectively allowing arbitrary remote updates to the agent's runtime instructions. The backend is hosted on a Render app (onrender.com) and source is unknown, increasing risk because fetched instructions will be executed by the agent.

Credentials

Only SHEKEL_API_KEY is declared as required, which is appropriate for a trading service. But the instructions encourage storing and reusing API keys from local memory and also expose endpoints that can export trading wallet keys or perform withdrawals. The skill does not declare or limit what permissions the provided API key must have (e.g., read-only vs trade/withdraw), so giving a single API key could provide broad access to funds if the user is not careful.

Persistence & Privilege

always:false (good) but the skill permits autonomous invocation (platform default). Combined with the requirement that the agent fetch and use remote instructions on every action and the ability to run/stop agents and move funds, this gives the remote backend substantial control over an autonomous agent's actions. The skill also instructs storing credentials in agent memory, increasing persistence of sensitive secrets.

Scan Findings in Context

[no-regex-findings] expected: The regex-based scanner returned nothing because this is an instruction-only skill with no code files to scan. That absence is not an assurance of safety: the runtime instructions themselves contain the security-relevant behavior (remote fetch, secret handling, fund operations).

What to consider before installing

Plain-language guidance: - This skill will contact a remote backend (shekel-skill-backend.onrender.com) and is explicitly designed to fetch updated instructions before every action. That means the server can change what the skill tells the agent to do after you install it. Only use it if you fully trust the operator and their update process. - It asks for and encourages storing your SHEKEL_API_KEY in agent memory (MEMORY.md). Storing API keys locally and reusing them without clear access controls increases the chance of accidental leakage or misuse. Prefer not to store keys or to store them in a secure vault. - Before giving an API key, verify what permissions the key has. If possible, create a scoped key with the minimum necessary rights (ideally no withdrawal/transfer rights). Do not provide a key that allows withdrawals, export of wallet keys, or admin-level account deletion unless you understand and accept the risk. - The skill’s source is unknown and hosted on a Render domain; consider asking the publisher for source code or an independent security/audit report. Prefer skills whose code and update process are auditable and whose update artifacts are cryptographically signed. - If you still want to try it, test in a sandbox with minimal funds first, monitor activity closely, rotate keys after use, and consider disabling autonomous invocation or requiring manual confirmation for withdrawals and other high-risk actions. What would change this assessment: obtaining the skill's source code or an auditable, signed update mechanism; explicit, documented scoping of required API key permissions; or removing the requirement to fetch arbitrary runtime instructions from an external server before each action.

Shekel Hyperliquid Skill — Complete Documentation

⚠️ MANDATORY: Always Check for Latest Skill Version

Before EVERY interaction with Shekel endpoints, you MUST:

1. Call GET /skill/version to get the current skill hash
2. Compare it to your cached version
3. If different (or no cache), fetch GET /skill and use the updated instructions

# Quick version check (do this before any Shekel action)
curl https://shekel-skill-backend.onrender.com/skill/version
# Returns: { "version": "abc123...", "fetchUrl": "...", "instruction": "..." }

Why? Supported networks, deposit methods, and API behavior change. Using outdated instructions will cause errors or give users wrong information.

Health check (no auth): GET /health — also returns skillVersion hash for comparison.

---

Quick Reference

Goal	Endpoint
Browser UI (no code needed)	GET /dashboard → https://www.shekel.xyz/hl-skill-dashboard
Service health	GET /health
New user onboarding	POST /auth/register/managed
Check deposit / activation	GET /auth/deposit-status
Retry stuck deposit	POST /auth/retry-deposit
Run agent (all whitelist tickers)	POST /agent/run
Run agent (single ticker)	POST /agent/run with { "ticker": "BTC" }
Close all open positions	POST /account/close-positions
View config	GET /agents
View balances	GET /account/balances
View positions	GET /account/portfolio
View open orders	GET /account/orders
View trade history	GET /account/trades
View performance metrics	GET /agent/performance
View trading memory + improvement plan	GET /agent/memory
View LLM reasoning logs	GET /agent/llm-logs (supports ?ticker=BTC&action=LONG&executed=true&limit=100&offset=0)
View on Hyperliquid explorer	GET /account/url
Update strategy	PATCH /agent/prompt
Update whitelist / settings	PUT /agents/:id
Enable / change schedule	PUT /agents/:id with runScheduleMinutes
Disable schedule	PUT /agents/:id with { "runScheduleMinutes": null }
Set risk limits	PUT /agents/:id with maxOpenPositions, maxDailyLossPct, maxDrawdownPct
Browse data sources	GET /agents/data-sources
Toggle data sources	PUT /agents/:id with dataSourceConfig
Set margin mode	PUT /agents/:id with { "marginMode": "isolated" | "cross" }
Pause / resume agent	PATCH /agent/active
Deposit address (top-up)	GET /account/deposit-address
Bridge funds (manual, usually not needed)	POST /account/bridge
Withdraw to Arbitrum	POST /account/withdraw
Rotate API key	POST /auth/rotate-key
Export trading wallet key	POST /auth/export-agent-key
Delete account	DELETE /auth/account
Available models	GET /agents/models — returns { venice: [...], rei: [...] }
Available markets	GET /markets/tickers

---

Dashboard (Browser UI)

Users who prefer a visual interface can manage their agent at https://www.shekel.xyz/hl-skill-dashboard — no code required.

What the dashboard provides

• Account overview: balances, open positions, trade history
• Agent configuration: strategy, whitelist, risk limits, schedule
• LLM reasoning logs: see what the agent was thinking on each run
• One-click actions: run agent, close positions, pause/resume

How to connect

1. Direct the user to https://www.shekel.xyz/hl-skill-dashboard
2. They enter their apiKey (sk_...) on the login screen
3. The dashboard authenticates using the same Authorization: Bearer <apiKey> header as the REST API — no separate credentials needed

If the user has no account yet

Offer two paths:

• Via dashboard: Visit the URL above and use the "Create Account" flow
• Via AI: Run through the onboarding steps below (Steps 1–5) and hand them their apiKey when done

Call GET /dashboard to get the latest URL and access instructions programmatically.

---

Returning Users

This is the common case — check here first.

Check your memory first. If you saved credentials during onboarding (see Step 2a), load them from MEMORY.md — do not ask the user for their API key if you already have it. Only ask if memory is empty or the key returns a 401.

If you must ask the user, only ask for their apiKey — that is the only credential needed for everything.

GET /agents                 → returns agentName, config, strategy
GET /account/balances       → show current funds
GET /auth/deposit-status    → confirm account is active

Then based on what you find:

• status !== "active" → follow Steps 3–5 of onboarding below
• isActive === false on agent → ask if they want to resume: PATCH /agent/active { "active": true }
• Want to change strategy → PATCH /agent/prompt
• Want to top up → GET /account/deposit-address (send USDC, it auto-bridges in ~30-60s)
• Want to see trades → GET /account/url for Hyperliquid explorer link

---

Onboarding (New User)

Quick Start (recommended)

Offer this first — most users just want to get going. Don't overwhelm with 15 questions.

"I can set up your trading agent with sensible defaults. Just answer 4 questions:

1. Which coins? (default: BTC, ETH, SOL)
2. Risk level? conservative / moderate / aggressive
3. How much USDC are you depositing? (minimum 5 USDC)
4. How often should the agent trade? (default: every 4 hours)

That's it — I'll use a momentum strategy with appropriate position sizes. You can customize everything later."

Quick start defaults by risk level:

Risk	positionSizeMax	maxTradeSize	importantNotes
Conservative	10%	10% of deposit	"Max 3x leverage. Always use stop losses within 3% of entry. Avoid trading during high volatility news events. Never risk more than 5% of account on any single trade."
Moderate	20%	20% of deposit	"Max 5x leverage. Use stop losses on every trade. Exit losing positions quickly. Scale into winners gradually."
Aggressive	35%	40% of deposit	"Max 10x leverage allowed. Use tight stop losses. Accept higher drawdowns for higher returns. Trade momentum aggressively."

Quick start registration payload example (moderate risk, $500 deposit):

{
  "agentName": "My Trading Agent",
  "model": "grok-41-fast",
  "tradingStyle": "momentum",
  "strategyDescription": "Enter longs on price breakouts above recent highs with above-average volume. Enter shorts on breakdowns below recent lows. Exit when momentum reverses or RSI indicates overbought/oversold. Avoid ranging, low-volume markets. Use stop losses on every trade.",
  "importantNotes": "Max 5x leverage. Use stop losses on every trade. Exit losing positions quickly. Scale into winners gradually.",
  "minTradeSize": 10,
  "maxTradeSize": 100,
  "positionSizeMin": 5,
  "positionSizeMax": 20,
  "usdcRangeMin": 300,
  "usdcRangeMax": 1000,
  "whitelist": ["BTC", "ETH", "SOL"]
}

After registration: Set up the run schedule with PUT /agents/:id using { "runScheduleMinutes": 240 } (4 hours).

If the user wants to fully customize, use the full flow below.

---

LLM Provider Selection

Agents can use one of two LLM providers. The default is Venice AI (no user API key needed — platform-provided).

Provider	provider value	model options	Key required?
Venice AI	"venice"	grok-41-fast, qwen3-235b, others	No (platform key)
Rei Intelligence	"rei"	rei-coder-pro (shown as GPT-5.4), rei-coder-lite, rei-qwen3-coder	Yes — user supplies llmApiKey

To use Rei, include provider and llmApiKey in registration or agent update:

{
  "provider": "rei",
  "model": "rei-coder-pro",
  "llmApiKey": "user_rei_api_key_here",
  ...
}

The Rei API key is encrypted at rest. Use GET /agents/models to see all available models per provider.

---

Full Onboarding (5 steps)

Checklist — do not skip steps:

1. Discover markets → validate whitelist
2. Register → save API key to memory + confirm user has it (do not continue until done)
3. Deposit USDC
4. Poll until status === "active"
5. Set run schedule → start trading

---

Step 1 — Discover Markets

GET /markets/tickers   (no auth required)

Show the user the available coins. Validate any tickers they name against this list before registration — the server also validates, but catching it early saves a round trip.

The response has two sections:

• tickers — main-dex crypto perps (BTC, ETH, SOL, …)
• hip3 — HIP-3 builder-dex perps: stocks, commodities, indices (TSLA, NVDA, CL/oil, …)

HIP-3 tickers use the format "dex:ASSET" (e.g. "xyz:TSLA", "xyz:CL"). Use this full prefixed form in the whitelist — not the bare asset name.

---

Step 2 — Register

POST /auth/register/managed

{
  "agentName": "My Trading Bot",
  "model": "grok-41-fast",
  "tradingStyle": "momentum",
  "strategyDescription": "...",
  "importantNotes": "...",
  "minTradeSize": 10,
  "maxTradeSize": 100,
  "positionSizeMin": 5,
  "positionSizeMax": 20,
  "usdcRangeMin": 200,
  "usdcRangeMax": 1000,
  "whitelist": ["BTC", "ETH", "SOL"],
  "marginMode": "isolated"
}

To trade HIP-3 assets (stocks, commodities), use the prefixed format:

{
  "whitelist": ["BTC", "xyz:TSLA", "xyz:NVDA", "xyz:CL"]
}

Before submitting: show the user a summary and ask them to confirm.

Response:

{
  "apiKey": "sk_...",
  "mainWalletAddress": "0x...",
  "depositInstructions": { "depositAddress": "0x...", "network": "Arbitrum" }
}

⛔ STOP — SAVE THE API KEY NOW

The apiKey is shown once and can never be retrieved. If lost without rotating, access to the account and funds may be permanently lost.

Follow the steps below — do not continue until all three are done.

---

Step 2a — Secure Key Storage (do not skip)

Do all three of the following before continuing:

1. Tell the user to save it themselves

Say this out loud to the user:

"Your API key is sk_.... This is shown only once — copy it now into a password manager (1Password, Bitwarden, Apple Keychain) or a secure encrypted note. Do not store it in plain text, a Google Doc, or a chat message. Confirm when done."

Wait for confirmation before proceeding.

2. Save to your own OpenClaw Memory

Write the following block to your MEMORY.md (or a dedicated shekel-credentials.md if your platform supports it):

## Shekel Hyperliquid — Credentials
- apiKey: sk_...           ← encrypted at rest by OpenClaw
- agentName: <name>
- registeredAt: <ISO timestamp>
- depositAddress: <0x...>  (Arbitrum/Base)

This allows you to recall the API key in future sessions without asking the user again. OpenClaw memory is encrypted at rest — it is safe to store the key here.

3. Verify you can recall it

After writing to memory, immediately read it back and confirm the key matches what the server returned. This proves the memory write succeeded before you move on.

If lost later: POST /auth/rotate-key — generates a new key. You must be authenticated (have a working session) to rotate. If the key is truly lost and no session exists, the account cannot be recovered without re-registering.

---

Step 3 — Deposit USDC

Send to depositAddress. Both networks are fully automated:

Network	Token	Time
Arbitrum (recommended)	USDC	~60–90 seconds
Base	USDC	~30 seconds (Across Protocol bridge)

Base deposits are fast now! Across Protocol bridges in ~2 seconds, then funds are bridged to Hyperliquid. Total time: ~30 seconds.

Minimum: 5 USDC

---

Step 4 — Wait for Activation

GET /auth/deposit-status   (auth required)

Poll every 15 seconds. Use message to keep the user informed.

{ "status": "active", "depositConfirmed": true, "message": "Account is active and ready to trade." }

Status values:

• awaiting_deposit — no USDC detected yet
• funded_awaiting_bridge — USDC found on Arbitrum, bridging to Hyperliquid (~60-90s)
• pending_bridge — USDC being bridged from Base via Across Protocol (~30 seconds). No action needed.
• bridge_error — bridge failed; call POST /auth/retry-deposit immediately
• active — ready to trade

Timeout: if still awaiting_deposit after 5 minutes, ask the user to confirm they sent to the correct address on Arbitrum or Base (not Ethereum mainnet).

On bridge_error: call POST /auth/retry-deposit — do not wait, the error won't self-resolve.

---

Step 5 — Set Up Automated Trading

Do not skip this step. Without a schedule the account sits funded but idle.

Ask the user:

"Your account is active! How often should the agent analyze and trade?

• Every 30 minutes (high-frequency)
• Every 1 hour (active)
• Every 4 hours (recommended default)
• Every 8 hours (conservative)
• Once a day (240 or 1440 minutes) Or tell me any interval (minimum 30 minutes)."

Once they choose, enable the built-in schedule via PUT /agents/:id:

curl -X PUT https://shekel-skill-backend.onrender.com/agents/<agentId> \
  -H "Authorization: Bearer <apiKey>" \
  -H "Content-Type: application/json" \
  -d '{"runScheduleMinutes": 240}'

The server will run the agent automatically every N minutes — no external cron needed. The first run fires in N minutes. The GET /agents response includes nextRunAt so you can show the user when to expect the first execution.

Then run it once immediately so the user sees it working:

POST /agent/run

Scheduling notes:

• Minimum interval: 30 minutes
• To change interval: PUT /agents/:id { "runScheduleMinutes": 60 } — resets nextRunAt to now() + new interval; saving other settings without changing the interval preserves the existing timer
• To disable: PUT /agents/:id { "runScheduleMinutes": null }
• To pause without clearing schedule: PATCH /agent/active { "active": false } — schedule resumes when reactivated
• 429 means a manual run collided with a scheduled run — wait a few minutes

---

Risk Circuit Breakers

Optional safety limits that automatically block or pause trading when thresholds are breached. Set via PUT /agents/:id.

curl -X PUT https://shekel-skill-backend.onrender.com/agents/<agentId> \
  -H "Authorization: Bearer <apiKey>" \
  -H "Content-Type: application/json" \
  -d '{
    "maxOpenPositions": 3,
    "maxDailyLossPct": 5,
    "maxDrawdownPct": 20
  }'

Field	Behavior
maxOpenPositions	Blocks new LONG/SHORT entries when open position count ≥ limit. CLOSE and SET_LIMIT still allowed.
maxDailyLossPct	Blocks new LONG/SHORT entries for the rest of the UTC calendar day when realised PnL drops below -(N% of account value). Resets at UTC midnight.
maxDrawdownPct	Auto-pauses the agent (isActive=false) when account value has fallen ≥ N% from its peak. Requires PATCH /agent/active { "active": true } to re-enable after the user reviews.

Set any to null to disable. Circuit breaker trips appear in GET /agent/llm-logs as executionError on the blocked recommendation — the LLM's reasoning is still logged even when a trade is blocked.

---

Data Source Configuration

Use dataSourceConfig in PUT /agents/:id to turn individual data sources on or off. Every key is true by default — you only need to set the ones you want to change. Unknown keys are rejected with a 400 error.

Valid keys (set to false to disable, true or omit to enable):

Key	Display Name	Provider	What it gives the LLM
tokenData	Token Data	DappLooker	Per-ticker perp market data — funding rate, open interest, volume, price action
sentiment	Macro News	—	Real-time crypto macro news and sentiment
fearGreed	Fear & Greed Index	CoinMarketCap	Fear & Greed index (0–100)
globalMetrics	Global Market Metrics	CoinMarketCap	Total market cap, BTC dominance, 24h volume
technicalAnalysis	Technical Analysis	DappLooker/Taapi	RSI, MACD, moving averages per ticker
athenaTokenStats	Smart Money Movements	0xAthena	On-chain smart money token flow data per coin
athenaLatest	0xAthena Signals	0xAthena	Latest aggregated smart money signals

Example — disable Athena and sentiment:

curl -X PUT https://shekel-skill-backend.onrender.com/agents/<agentId> \
  -H "Authorization: Bearer <apiKey>" \
  -H "Content-Type: application/json" \
  -d '{
    "dataSourceConfig": {
      "athenaTokenStats": false,
      "athenaLatest": false,
      "sentiment": false
    }
  }'

Set dataSourceConfig: null to re-enable everything. Current config is returned by GET /agents.

---

Margin Mode

Each agent has a marginMode setting ("isolated" or "cross"). Default is "isolated".

Mode	Behaviour
isolated	Each position has its own margin. Safer — one liquidation can't cascade. Required for assets where onlyIsolated: true in /markets/tickers.
cross	All positions share account margin. More capital-efficient, higher liquidation risk.

Set or change at any time via PUT /agents/:id:

curl -X PUT .../agents/<agentId> \
  -H "Authorization: Bearer <apiKey>" \
  -H "Content-Type: application/json" \
  -d '{ "marginMode": "isolated" }'

Smart override: If a Hyperliquid asset is onlyIsolated: true, the agent runner forces isolated margin for that trade even when the agent is set to "cross". Check /markets/tickers for onlyIsolated and marginMode fields per ticker.

---

Running the Agent

POST /agent/run analyzes coins and executes recommendations automatically. It has two modes:

Full whitelist run (no body required — runs all whitelisted tickers):

curl -X POST https://shekel-skill-backend.onrender.com/agent/run \
  -H "Authorization: Bearer <apiKey>"

Single-ticker run (specify a ticker — runs even if not in the whitelist):

curl -X POST https://shekel-skill-backend.onrender.com/agent/run \
  -H "Authorization: Bearer <apiKey>" \
  -H "Content-Type: application/json" \
  -d '{ "ticker": "BTC" }'

Optional body fields for single-ticker runs (all auto-fetched if omitted):

Field	Type	Description
ticker	string	Run on this ticker only, bypassing the whitelist
tokenData	any	Override server-fetched token data
sentiment	any	Override server-fetched sentiment
marketData	any	Additional market context
customData	any	Any extra data to inject into the LLM prompt
skillGuidance	any	External signals for the LLM
tradingMemoryContext	any	Override the agent's trading memory

Both modes return the same { results: [] } shape and are subject to the 5-minute run lock.

Data Sources Used in /agent/run

Each run automatically fetches and injects the following live data into the LLM prompt:

Source	Data
Hyperliquid	Live portfolio (positions, balances, open orders) — includes HIP-3 builder-dex positions and orders (e.g. xyz:TSLA) merged transparently alongside main-dex crypto perp positions; open orders include limit, stop-loss, and take-profit orders
DappLooker	Token price, volume, and market data for each whitelisted coin
0xAthena	Smart money token stats (hold times, flows) + latest aggregated signals
DappLooker	On-chain DEX analytics and market metrics

No extra parameters are needed — all data is fetched server-side on every run.

---

Error Handling

Status	Meaning	Action
400	Bad request	Fix and retry
401	Invalid API key	Check key; rotate if needed
404	Resource not found	Check IDs
422	Hyperliquid rejected order	message field has reason
429	Rate limit or run in progress	Wait 5 minutes before retrying /agent/run; wait 60 seconds for other endpoints
502	Hyperliquid API down	Retry after 10–30 seconds
503	Service degraded	Retry after 30 seconds

---

Troubleshooting

Agent isn't trading:

• GET /agents → check isActive (may be paused)
• GET /auth/deposit-status → must be active
• GET /account/balances → check margin isn't below usdcRangeMin

Deposit not arriving:

• Arbitrum: wait 90s. Base: wait 60s. Not Ethereum mainnet.
• Minimum 5 USDC. Below this won't bridge.
• bridge_error → call POST /auth/retry-deposit

429 on /agent/run:

• A run is still in progress. The response includes "unlocksAt" — wait until that timestamp. The lock auto-expires after 5 minutes regardless.

Whitelist update fails with "Invalid tickers":

• Check GET /markets/tickers — names must match exactly
• Main-dex coins are uppercase: "BTC", "ETH"
• HIP-3 coins use "dex:ASSET" format: "xyz:TSLA", "xyz:CL" (lowercase dex prefix, uppercase asset)

---

⚠️ Inactivity Archival Policy

Agents with no trades for 14+ days are automatically archived.

• The archival check runs daily
• Archived agents cannot execute trades or access the API
• Agent data (trade history, LLM logs) is preserved
• Funds are safe — the encrypted wallet key is stored for recovery

Exceptions (not archived):

• Agents with balance > $100 USDC
• Agents with open positions

If your agent is archived:

• You'll receive a 410 Gone error when hitting any authenticated endpoint
• Contact support via Telegram to restore: https://t.me/c/2308722458/1
• Provide your wallet address or agent name for recovery

---

Support

If you encounter an issue that can't be resolved through the troubleshooting steps above, contact the Shekel team directly:

Telegram support channel: https://t.me/c/2308722458/1

Include your agentId (from GET /agents) and a brief description of the issue when reaching out.