Hyperliquid names one official SDK and a short list of community ones. Everything else you find searching GitHub for a Hyperliquid trading bot needs its own check, because real malware campaigns have used fake GitHub repos and Hyperliquid-named packages to find victims. Here is what Hyperliquid itself points to, what has actually gone wrong, and how to lower the risk if you run someone else’s code anyway.
November 16, 2026
14
min to read
The short answer
Hyperliquid names one official SDK, for Python, at github.com/hyperliquid-dex, and a short list of community SDKs its own docs link to by name. Everything else on GitHub needs its own check: real malware has shipped as fake GitHub repositories and npm packages, some of them typosquatting the names of trusted developers.
Hyperliquid’s own organization on GitHub, github.com/hyperliquid-dex, held 11 public repositories when we read it on November 16, 2026: the Python SDK, the software that runs a validator or a regular node, the audited bridge contracts, two stats repos, a Rust SDK, a folder of TypeScript examples, and a few smaller infrastructure tools. Sitting in that org is a real signal. It is not the only one worth checking.
Hyperliquid’s developer docs list SDKs on their own API page, and only one of them lives in Hyperliquid’s own organization: the Python SDK, at the same address. For every other language, the page sends you somewhere else entirely. Its Rust SDK is described as “written by community member, Infinite Field,” hosted at github.com/infinitefield/hypersdk, a different GitHub account altogether, not the hyperliquid-rust-sdk repo that happens to sit inside hyperliquid-dex itself (475 stars, last updated October 2025, and not the one the current docs page points to). Its two TypeScript options, by developers named nktkas and nomeida, are community-maintained and live outside the org too. CCXT, a library that speaks to many exchanges at once, gets a third-party mention of its own.
The lesson sits in that mismatch. An address inside the official org is worth something, but it is not the whole test, and an address outside it is not automatically a problem; what matters is whether today’s docs page, read on the date you are checking, actually names the repo in front of you.
We read each repo’s README directly rather than take someone’s word for it. A few things stand out.
account_address, your main wallet’s public address, and secret_key, a private key to sign with. Right under that, an optional step tells you to generate a fresh key on app.hyperliquid.xyz/API instead, and to put that key in secret_key while still putting your main wallet’s public address, “not the API wallet,” in account_address. Getting that one line backwards is an easy way to confuse the two.gpg and check it against a signature file before you run the binary. That is the standard Hyperliquid holds itself to for its own infrastructure: a key you can check a download against, not a file you just trust.None of this is a security feature built for GitHub browsers; it is how a careful team documents its own code. Reading it costs a few minutes and tells you more than a star count ever will.
This is not a hypothetical. Kaspersky’s security research team has documented two campaigns that show the two different shapes the risk takes: a fake repo built to look real, and a real, popular package turned hostile after the fact.
Kaspersky’s GitVenom report, published February 24, 2025, found hundreds of fake GitHub repositories posing as useful tools, including an Instagram automation tool, a Telegram bot for managing Bitcoin wallets, and a cheat for the game Valorant. Each one carried a convincing, well-written README, and a timestamp file rewritten every few minutes inflated the number of commits, so the repo looked busy and maintained. The code itself hid its payload differently by language: encrypted in Python, Base64-encoded in JavaScript, buried in a Visual Studio project file in C, C++ and C#.
Once installed, victims got a Node.js stealer that gathered saved credentials, cryptocurrency wallet data and browsing history and shipped it out over Telegram, plus backdoors (AsyncRAT and Quasar) and a clipboard hijacker that swapped a copied wallet address for one the attackers controlled. One of those attacker addresses alone took in about 5 BTC, roughly $485,000 at the time of Kaspersky’s research, in a single payment in November 2024. Kaspersky observed infection attempts worldwide.
Kaspersky’s report on the Shai-Hulud worm, published September 25, 2025, describes a different shape entirely: no fake repo to spot, because the packages were already real, already popular, and already installed. The worm read a victim’s npm and GitHub access tokens, used a secrets-scanning tool with AWS and GCP helpers to grab whatever credentials it could reach, and pushed what it found to a new repository, named “Shai-Hulud,” created right on the victim’s own GitHub account. Then it modified that same victim’s 20 most-downloaded packages to carry the same payload, bumped each one’s version number and republished it, so every newly infected machine turned around and poisoned the packages other developers would install next. Kaspersky counted more than 500 npm packages caught in it, including one, @ctrl/tinycolor, with more than two million weekly downloads, and nine packages published under the security firm CrowdStrike’s own name.
Neither campaign targeted Hyperliquid by name, but together they are exactly the two risks a Hyperliquid trading bot’s dependencies carry: a repo dressed up to look like what you searched for, and a package you already trusted turning on you without you touching a line of your own code.
GitHub runs its own Advisory Database, and a search for “hyperliquid” malware there, on November 16, 2026, listed 10 malware advisories across 7 distinctly named npm packages. One of them is the clearest lesson in this whole guide: a package published as @nklkas/hyperliquid, a one-letter swap of nktkas, the handle behind the real, widely used TypeScript SDK named on Hyperliquid’s own docs page and published on npm as @nktkas/hyperliquid. Type the name from memory, or copy it from a search result instead of the docs page, and that single swapped letter is the entire difference between the real SDK and a package GitHub itself has flagged as malware.
Before anything else, check whether the package name, spelled exactly, appears on Hyperliquid’s own API docs page. A one-letter difference, as above, is the whole trick.
A repo inside github.com/hyperliquid-dex is a real signal, and one outside it is not automatically a problem: the community SDKs the docs recommend both live elsewhere. Either way, open the author’s other work and history; a name with nothing behind it but this one repo is a reason to slow down.
GitVenom’s repos had polished READMEs and fake, auto-updated commit history. Stars and forks can be bought or faked too. Open the files that actually trade or handle a key, and look for anything that runs automatically on install, downloads a separate binary, or calls itself a “loader.”
Shai-Hulud spread by republishing a slightly newer version of a package already on people’s machines. A pinned, lockfile-checked dependency cannot be swapped out from under you by a later update alone.
Whatever you decide to run, give it a Hyperliquid API wallet, never your seed phrase or your main private key. The next section covers exactly what that key can and cannot do.
Run anything new with no real funds anywhere near the process, on a machine that holds no other keys, before it ever touches mainnet.
Even careful vetting can miss something, which is why the key you hand to any Hyperliquid bot matters as much as the code itself. Hyperliquid’s own API page describes an API wallet, also called an agent wallet, as a key that “can perform actions on behalf of an account without having withdrawal permissions.” It can place orders, change leverage and move your own USDC between your own balances. It cannot send money to anyone else’s address, and it cannot approve another API wallet in your place; only your main wallet can do either of those.
Hyperliquid’s API wallet docs set the limits that make this a leash rather than an open door:
Put it into practice this way: give every piece of outside code its own named API wallet, never one shared with anything else, so a bad tool’s worst case is exactly that one key and nothing more. Remove it from the API page in the app the moment you stop trusting the code behind it, instead of leaving it live “just in case.” And keep the limit of this protection in view: an API wallet stops someone from draining your account to another address, but it can still lose your money through bad trades, because it can still trade. It caps the damage; it does not remove the risk of running code you have not read.
If you already ran something you now doubt, or you see activity you did not start, Hyperliquid’s own guidance for a compromised account is the same regardless of what caused it: stop using that address, move what is left to a wallet created fresh, and undo any smart-contract permission it was ever given, with the approval checker that page links to, before you trust that device again.
A few routes put a traceable name behind the code, instead of a GitHub username you have never heard of:
None of these removes every risk Hyperliquid itself carries, such as leverage or a market moving fast; our guide to whether Hyperliquid is safe covers those separately. They remove the specific risk this guide is about: code whose author you cannot name.
Ethy is an AI trading agent you ask before you trade, reached in a browser or app rather than cloned from a repository. You never download, run or trust someone else’s trading code to use it, which sidesteps the whole question this guide is about.
Under the hood, its Hyperliquid orders go through the kind of key this guide recommends. When you activate it, Ethy is set up to sign for your wallet from its servers, which is how automations run while you are away, and it creates its own API wallet, named ethy-perps, which your wallet approves. That API wallet places the orders; its private key is stored encrypted on Ethy’s servers rather than published anywhere, and it is renewed automatically rather than left to quietly lapse. Ethy is software for research and analysis, so it does not give investment advice, but it can read a chart on four timeframes, look into why a market moved, and test a rule on past data, before you decide whether any bot, including one you found on GitHub, is worth trusting with real money.
Hyperliquid names one official SDK, for Python, at github.com/hyperliquid-dex, which you can build a bot on top of yourself. It is not a ready-made bot; it is the toolkit the docs point to for talking to the exchange. Hyperliquid’s docs also name community SDKs for other languages, by author, which live outside that same GitHub organization.
Check its name against the SDK list on Hyperliquid’s own current API docs page, check who actually maintains it and for how long, and read the trading code itself rather than trust its README or its star count. Real malware campaigns have used polished READMEs and inflated activity to look trustworthy, so none of those surface signals proves anything on their own.
Yes. Kaspersky’s GitVenom report, though not aimed at Hyperliquid, documented fake repos, including one posing as a Telegram Bitcoin wallet bot, that stole wallet data and hijacked clipboards. GitHub’s own Advisory Database separately listed malware advisories for several npm packages named after Hyperliquid, including one that typosquats the handle of a real community SDK’s maintainer by a single letter.
An API wallet cannot withdraw or send funds to another address, per Hyperliquid’s own API documentation, so it blocks the most direct kind of theft. It can still place real trades, so malicious or broken code behind that key can still lose your money through bad orders. Treat it as a way to cap the damage, not a reason to skip vetting the code.
It is malicious code smuggled through a package your bot installs as a dependency, sometimes by compromising a package that was already legitimate, as Kaspersky’s Shai-Hulud report describes. A Hyperliquid bot written in JavaScript or TypeScript installs npm packages the same way any other project does, so the same risk applies to its dependencies, not only to the bot’s own code.
Match its exact name against Hyperliquid’s current docs page, check the author’s history beyond this one repo, read the actual trading code rather than the README, pin dependencies with a lockfile, and run it on testnet with an API wallet before any real funds are near it. Never your seed phrase or main private key, on any bot.
Trading involves risk, including the loss of your money. Ethy AI is a research and trading tool and does not give investment advice or recommendations. Backtests are simulations on historical data and do not predict future results.
Deploy your AI trading agent in one click.
