Skip to content

Security

What you have to trust.

Two things matter more than a badge: what the protocol structurally cannot do to you, and what has genuinely been reviewed. Both are below, including the parts with no published report.

Trust model

Start with what is structural.

An audit is a snapshot of code at one moment. These properties hold continuously, because they are enforced by Sui's object model and by parameters anyone can read.

  • Collateral is a user-owned object

    Perpetuals collateral lives in an Account object owned by your address — a transferable Sui object, not a credit line on Aftermath's books. Margin is isolated per market, so a loss in one market cannot reach collateral in another.

  • Liquidations are permissionless and partial

    Any address can call a liquidation; there is no privileged liquidator. A liquidation reduces the position to the market's maximum initial leverage rather than closing it, and pays three fees: a liquidation fee and a force-cancel fee to the liquidator, and an insurance-fund fee to that market's isolated fund.

  • Margin parameters change on a delay

    IMR and MMR are stored onchain. A change is written onchain before it takes effect and is time-delayed by a few days, so nobody can tighten margin requirements out from under an open position without warning.

  • Prices are aggregated, not single-sourced

    Oracle prices are aggregated across Pyth, Stork, Switchboard, and SEDA. The mark price used for margin and liquidation is the median of the book price, the funding price, and the TWAP price, so no single input decides a liquidation.

  • Bad debt has a defined waterfall

    If a liquidation leaves bad debt, the market's isolated insurance fund absorbs it first. If that is exhausted the loss is socialized, and auto-deleveraging is the last step. The order is fixed in the contracts, not decided during an incident.

  • Delegation cannot withdraw

    An agent wallet can perform all actions except withdrawing collateral and granting or revoking other agent wallets. Builder codes carry a user-approved maximum fee cap and can be revoked. Both are enforced in the onchain fee and permission logic.

  • Staked positions stay untouchable

    A farm position is a user-owned Sui object that only its owner can invoke. That is why auto-compounding is impossible on Aftermath farms: the vault does absolutely nothing with the funds you stake, because it cannot access them.

  • Parameters are readable by anyone

    Fees, margin requirements, insurance-fund balances, and vault limits are onchain state. Every mainnet package and object id is published, so you can read the live values instead of the documented ones.

Audits

Two reports, and their limits.

Two OtterSec reports are published. They are linked here in full, unedited, so you can read the findings rather than a summary of them.

Scope: what is not covered

The reports above cover Pools and afSUI liquid staking. They do not cover the rest of the protocol. There is no published third-party review for:

  • Perpetuals
  • Perpetuals vaults (afLP and permissionless vaults)
  • Farms
  • DCA
  • Limit orders
  • NFT AMM

That is a statement about what has been published, not a claim that the code is unreviewed internally. Treat unaudited contracts as unaudited contracts and size positions accordingly.

Reporting

How to report a vulnerability.

The honest version: there is one channel, and it is Discord. Anything else you may have seen described does not exist yet.

There is no bug bounty program and no published security contact address or PGP key. Do not send a report expecting a reward schedule or a triage SLA — neither is documented today.
  1. 01

    Open a Discord ticket

    Report in a ticket in the Aftermath Discord rather than posting details in a public channel. This is the same channel the docs direct API rate-limit requests to, so it is monitored.

  2. 02

    Include onchain evidence

    Include the affected package or object id, the transaction digests that demonstrate the issue, and the network. Mainnet ids are published on the contracts page.

  3. 03

    Coordinate disclosure

    Withhold public disclosure until the team has responded. Aftermath does not publish a disclosure window, so agree on one directly in the ticket.

Access

Geographic restrictions.

The protocol is a contract on a public chain; the interfaces that front it are not available everywhere.

OFAC-sanctioned countries
Blocked entirely
United Kingdom
No perpetuals terminal or account management
United States
No perpetuals terminal or account management

The legal documents govern in full. Note that the Terms of Service and the Privacy Policy currently name different operating entities; both are linked below, unmodified, rather than paraphrased here.

Risk

What can go wrong.

Each of these is documented by Aftermath. None of them is a reason not to use the protocol; all of them are reasons to size a position deliberately.

  • Leverage risk

    Perpetuals are margined products. An IMR of 0.1 means up to 10× leverage, and an MMR of 0.05 means liquidation at 20× effective leverage. Positions can be liquidated in full over time through repeated partial liquidations.

    Docs
  • Socialized loss and ADL

    If a market's insurance fund cannot cover bad debt, losses are socialized and profitable positions can be auto-deleveraged. A position can be closed by the protocol without your instruction.

    Docs
  • Vault risk

    The documentation lists strategy, market, smart contract, and curator risk for permissionless vaults, and says of the afLP vault plainly that it is not a risk-free yield product. Deposits are locked for four days, with force withdrawal available one day after a request sits idle.

    Docs
  • Curator risk

    A vault curator runs a strategy with pooled capital. The protocol caps the curator fee at 20% of profits, the lock at 60 days, and the market count at 12 — but it does not evaluate the strategy for you, and there is no close-vault function.

    Docs
  • Smart contract risk

    Every protocol here is code deployed onchain. Two of them have a published third-party review; the rest do not. Code that has been reviewed can still contain faults.

    Docs

FAQ

Frequently asked questions

Verify it yourself.

The AMM is open source and every mainnet package id is published. Read the code and the live parameters instead of taking this page's word for them.