Helix Finance rulebook v3
CurrentAdded EXCL-8 (malicious browser extensions). Adopted the SEAL Whitehat Safe Harbor agreement.
Assets in scope
4 assets
| ID | Asset | Type | Target |
|---|---|---|---|
| HLX-1 | Pool (proxy) | Smart contractEthereum | 0x4c1f2E7e…9A8b7C6d |
| HLX-2 | Liquidator | Smart contractEthereum | 0x8A9b0C1d…4E5f6A7b |
| HLX-3 | App Frontend only; no backend in scope. | Website / app | https://app.helix.example |
| HLX-4 | Pool (Arbitrum) | Smart contractArbitrum | 0x4c1f2E7e…9A8b7C6d |
Impacts & severity
Severity follows mechanically from the demonstrated impact.
- IMP-C1Direct theft of any user funds, whether at-rest or in-motion, other than unclaimed yieldMust demonstrate: Attacker-controlled address ends with assets previously owned by a user or the protocol, without their authorization.
- IMP-C2Permanent freezing of fundsMust demonstrate: User or protocol funds become irrecoverable by any party, including via upgrade or governance.
- IMP-C3Protocol insolvencyMust demonstrate: Liabilities exceed assets such that not all users can withdraw.
- IMP-H1Theft of unclaimed yieldMust demonstrate: Attacker obtains yield/rewards accrued to other users that had not yet been claimed.
- IMP-H2Temporary freezing of funds for at least 24 hoursMust demonstrate: Users cannot withdraw for ≥ 24h, recoverable without loss of principal.
- IMP-M1Smart contract unable to operate due to lack of token fundsMust demonstrate: A core function reverts for all users because the contract lacks required balance.
- IMP-M2Griefing (no profit motive for attacker, but damage to users or protocol)Must demonstrate: Demonstrated measurable loss to users/protocol with no attacker profit.
- IMP-L1Contract fails to deliver promised returns, but doesn't lose valueMust demonstrate: Returns deviate from documented behavior; principal unaffected.
Proof-of-Concept requirements
- POC-1Critical and High reports must include a runnable PoC (Foundry or Hardhat) against a mainnet fork demonstrating the claimed impact end-to-end.
- POC-2The PoC must state the fork block number and must never be executed against mainnet or public testnets.
- POC-3Funds-at-risk claims must reference on-chain balances at the stated fork block.
Testing restrictions
Allowed: Local mainnet fork (Foundry/Anvil, Hardhat) · Project-provided staging deployment
- TEST-1Testing on mainnet or public testnet deployed code is prohibited; use local forks.
- TEST-2Testing against pricing oracles or third-party smart contracts is prohibited.
- TEST-3Phishing, social engineering and physical attacks are prohibited.
- TEST-4Denial-of-service attacks against project infrastructure are prohibited.
Out-of-scope & exclusions
A closure as out of scope must cite one of these ids.
- EXCL-1Attacks that the reporter has already exploited themselves, leading to damage.
- EXCL-2Attacks requiring access to leaked keys or credentials.
- EXCL-3Attacks requiring access to privileged addresses (governance, admin, guardian) except where the privilege is obtainable by an unprivileged attacker.
- EXCL-4Incorrect data supplied by third-party oracles (oracle manipulation and flash-loan attacks remain in scope).
- EXCL-5Basic economic governance attacks (e.g. 51% attacks).
- EXCL-6Lack of liquidity impacts and best-practice recommendations without demonstrated impact.
- EXCL-7Centralization risks acknowledged in the published audit reports.
- EXCL-8Issues in the frontend (HLX-3) that require the user to install a malicious browser extension.
Policies
The first report that identifies a root cause is eligible. Later reports sharing the same root cause are duplicates, even when they describe a different attack path, unless they demonstrate a materially higher impact.
Issues listed in published audits or recorded as known-issue commitments before the submission timestamp are ineligible. Commitments may be revealed during a dispute.
Public disclosure requires approval from both parties after the fix is deployed and verified. Default embargo: 30 days after fix.
Either decision may be escalated to an independent platform mediator within 7 days of the project's final decision. The mediator publishes a reasoned decision to both parties.
Rewards are paid in USDC within the payout SLA after retest passes or after a Won't Fix decision. A Won't Fix decision never cancels an earned reward. The amount equals the locked reward calculation.
Adopted. During an active exploit, whitehats are pre-authorized to rescue funds under the SEAL agreement, for 10% of recovered assets up to $1,000,000. Adoption record: https://safeharbor.securityalliance.org