agentleFS
galleonlabsgalleonlabs/hypergrok-trading-desk/skills/hyperliquid-setup/SKILL.md

hyperliquid-setup

Skill418 starsChanged 12 days agoMITView on GitHub

Prepare the desk computer to work with Hyperliquid - install the SDKs, pick testnet or mainnet, verify connectivity, and (only when the user asks) provision a trade-only API wallet through the secure secret store and verify it is approved. Use during desk setup, when moving between research, testnet and mainnet levels, when a key is rotated, or when any Hyperliquid call fails with an environment problem.

  • Installs packages

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never ask for the seed phrase or the main wallet's key.

    line 84
  2. Very specific:

    If the desk's Grok Bot setup exposes secrets to the computer as environment variables, scripts read HYPERLIQUID_PRIVATE_KEY from the …

    line 90
  3. Distinctive:

    Every signing snippet on the desk uses the same loader, which reads the environment first, then that file, and never prints the key:

    line 98
  4. Distinctive:

    Run by the Execution Trader.

    line 126

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Hyperliquid setup

Sections 1-3 are read-only and safe to run at any time. Section 4 provisions a key and is done only when the user asks to move to a testnet or mainnet desk. Section 5 is the readiness check the Execution Trader runs before its first send.

1. Networks

MainnetTestnet
RESThttps://api.hyperliquid.xyzhttps://api.hyperliquid-testnet.xyz
WebSocketwss://api.hyperliquid.xyz/wswss://api.hyperliquid-testnet.xyz/ws
Apphttps://app.hyperliquid.xyzhttps://app.hyperliquid-testnet.xyz
Faucet-https://app.hyperliquid-testnet.xyz/drip

Reads (/info) need no key. Writes (/exchange) need a signing key and are signed differently per network, so a key and its network go together. The desk records the network in /workspace/trading-desk/desk.md and in the environment.

Testnet is the default. Nothing on the desk selects mainnet unless HYPERLIQUID_NETWORK=mainnet is set on purpose.

2. Install

Python (used by most snippets on the desk):

python3 --version                                  # 3.9+ required, 3.11+ preferred
python3 -m venv ~/.hypergrok/venv
source ~/.hypergrok/venv/bin/activate
python3 -m pip install "hyperliquid-python-sdk==0.24.0"
python3 -c "import hyperliquid, eth_account; print('ok', hyperliquid.__name__)"

Activate the same environment with source ~/.hypergrok/venv/bin/activate in each new shell before SDK snippets. The public Python tools shipped here need no SDK or activation. These are the versions reviewed on 2026-10-25; review API and snippet compatibility before upgrading them.

Optional TypeScript (if the user prefers TS or wants the WebSocket client from @nktkas/hyperliquid):

node --version                                      # 22.12+ required by @nktkas/hyperliquid (ESM-only)
mkdir -p /workspace/trading-desk/ts && cd /workspace/trading-desk/ts
[ -f package.json ] || npm init -y >/dev/null
npm pkg set type=module >/dev/null                  # so .js files can use import and top-level await
npm i @nktkas/hyperliquid@0.33.3 viem@2.57.11 decimal.js@10.6.0
node -e "import('@nktkas/hyperliquid').then(m=>console.log('ok', Object.keys(m).length))"

curl and jq are enough for every read in hyperliquid-market-data and hyperliquid-account; install jq if missing.

3. Connectivity check (no key)

for base in https://api.hyperliquid.xyz https://api.hyperliquid-testnet.xyz; do
  printf '%s -> ' "$base"
  curl -sS -m 10 -X POST "$base/info" -H 'Content-Type: application/json' \
    -d '{"type":"allMids"}' | jq -r 'if type=="object" then "ok, \(length) mids, BTC \(.BTC)" else "unexpected: \(.)" end'
done

Both lines should read ok. Record the time and result in the desk record. If one network fails, the desk is not blind on the other; say which.

4. API wallet (only when the user asks to trade)

Before creating or provisioning any key, verify and record native out-of-band approval coverage for the actual SDK, script and HTTP send paths, per desk-operating-model and SETUP.md section 7. If it cannot be verified, continue research without a key. An existing key or chat approval does not remove this prerequisite.

An API wallet (Hyperliquid also calls it an agent wallet) is a separate key the user authorises to sign trading actions for their account. It can place, modify and cancel orders and change leverage and margin; it cannot withdraw to Arbitrum, cannot send USDC or tokens to another wallet, and cannot approve other agents or builders (those need the main wallet's signature). It can still move funds inside the user's own account (perp to spot and across dexs via agentSendAsset, into sub-accounts and vaults, and it can spend USDC on reserveRequestWeight), all of which the desk forbids by rule. That is the only kind of key that ever reaches the desk computer.

Why: all of the user's Bots share one computer, so anything on it is readable by every Bot. A trade-only key bounds the damage; a seed phrase or main-wallet key would not.

4.1 The user creates it in the Hyperliquid app

Guide the user; do not do it for them (it involves their wallet):

  1. Open the app for the chosen network (https://app.hyperliquid.xyz or https://app.hyperliquid-testnet.xyz), connect the main wallet, open the API page (/API, under More in the app's navigation).
  2. Generate a new API wallet, give it a name such as hypergrok-desk, choose a validity period (a few months is sensible; it can be revoked at any time), and Authorize. The app shows the API wallet's private key once. The API wallet's address is shown next to it.
  3. On testnet, fund the account from the faucet (/drip): it gives 1,000 mock USDC, but only to an address that has deposited on mainnet at some point (email/Privy logins get different addresses per network; export the mainnet wallet into a wallet extension and connect that to testnet). Testnet USDC is not real.

Never ask for the seed phrase or the main wallet's key. If the user offers it, decline and point back to this step.

4.2 The user provides the key through the secure secret store

Grok Bot has a secure secret card for exactly this. Ask the user to add a secret named HYPERLIQUID_PRIVATE_KEY with the API wallet's private key. Never ask them to paste it in chat and never write it to a file under /workspace.

If the desk's Grok Bot setup exposes secrets to the computer as environment variables, scripts read HYPERLIQUID_PRIVATE_KEY from the environment. If it does not, the fallback is a file the user creates while in control of the computer:

mkdir -p ~/.hyperliquid && chmod 700 ~/.hyperliquid
# the user pastes the key into this file themselves, then:
chmod 600 ~/.hyperliquid/api-wallet.key

Every signing snippet on the desk uses the same loader, which reads the environment first, then that file, and never prints the key:

import os
def load_key():
    k = os.environ.get("HYPERLIQUID_PRIVATE_KEY")
    if not k:
        p = os.path.expanduser("~/.hyperliquid/api-wallet.key")
        if os.path.exists(p):
            k = open(p).read().strip()
    if not k:
        raise SystemExit("no API wallet key available - see hyperliquid-setup section 4")
    return k

4.3 Non-secret configuration

Set these for the desk computer's shell (for example in ~/.bashrc, and mirror them in desk.md):

export HYPERLIQUID_NETWORK=testnet                 # or mainnet, deliberately
export HYPERLIQUID_ACCOUNT_ADDRESS=0xYourMainAccount   # the account the API wallet acts for, NOT the API wallet's address

The account address is public. Reads use it; the Execution Trader passes it as account_address when signing with the API wallet.

5. Readiness check (before the first send)

Run by the Execution Trader; read-only; proves the key belongs to the account on this network without sending anything.

import os, time
import eth_account
from hyperliquid.info import Info
from hyperliquid.utils import constants

def load_key():
    k = os.environ.get("HYPERLIQUID_PRIVATE_KEY")
    if not k:
        p = os.path.expanduser("~/.hyperliquid/api-wallet.key")
        if os.path.exists(p):
            k = open(p).read().strip()
    if not k:
        raise SystemExit("no API wallet key available - see hyperliquid-setup section 4")
    return k

network = os.environ.get("HYPERLIQUID_NETWORK", "testnet")
base = constants.MAINNET_API_URL if network == "mainnet" else constants.TESTNET_API_URL
account = os.environ["HYPERLIQUID_ACCOUNT_ADDRESS"]
agent_addr = eth_account.Account.from_key(load_key()).address   # derived locally, key never printed

info = Info(base, skip_ws=True)
role = info.user_role(agent_addr)                       # {"role": "agent", "data": {"user": "0x..."}} when approved
agents = info.extra_agents(account)                     # named agents: [{"address","name","validUntil"}, ...]
named = next((a for a in agents if a["address"].lower() == agent_addr.lower()), None)
acts_for_account = role.get("role") == "agent" and role.get("data", {}).get("user", "").lower() == account.lower()
unexpired = named is None or named["validUntil"] > time.time() * 1000
approved = acts_for_account and unexpired               # an unnamed agent is proved by userRole alone
state = info.user_state(account)

print(f"network={network} account={account}")
print(f"api_wallet={agent_addr} role={role.get('role')} approved_for_account={approved}")
for a in agents:
    if a["address"].lower() == agent_addr.lower():
        print(f"valid_until_ms={a.get('validUntil')} name={a.get('name')}")
print(f"account_value={state['marginSummary']['accountValue']} positions={len(state['assetPositions'])}")

Ready means: role=agent, approved_for_account=True, validUntil in the future (when the agent is named and listed), and account_value is what the user expects on this network. Anything else: stop and report which line disagrees.

6. Rotation and revocation

  • Revoke: the user removes the API wallet on the app's /API page. Sends from the desk fail immediately afterwards.
  • Rotate: revoke, create a new API wallet (4.1), replace the secret (4.2), rerun the readiness check. Journal the time.
  • Suspected misuse: revoke first, investigate second (desk-incident-response playbook G).

Pitfalls

  • Using the API wallet's address as HYPERLIQUID_ACCOUNT_ADDRESS. Reads on the agent's address show an empty account; the account is the main wallet's address.
  • Signing for mainnet with a testnet key or vice versa. The signature encodes the network; the exchange rejects the mismatch, but the desk should never get that far.
  • Installing the SDK into a temporary environment or forgetting to activate it in a new shell. Keep ~/.hypergrok/venv; re-run section 2 if that directory is lost after a computer refresh.
  • Info(...) performs network calls on construction; pass skip_ws=True unless you want the WebSocket thread.
  • Printing environment variables while debugging. Redact anything that could be the key.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.