Loops HouseStead

NEAR HACKATHON
Team
1 member- SAOwner
Samuel Chuku
Overview
Problem: when an Aave position’s health factor drops below 1.0, a liquidator repays part of the debt and takes a bonus (5% for WETH) out of the borrower’s collateral. Unless the borrower acts in time, they pay that penalty.
Stead is a liquidation rescue market on NEAR. A NEAR contract holds the borrower’s Aave V3 position on Ethereum through chain signatures: only the contract can sign for that address, from a fixed allow-list of calls. An OutLayer program proves the position’s health from Ethereum storage proofs inside a TEE, and every decision rests on that verified reading. When health falls below 1.15, steward agents compete to lend the position the collateral it needs: each bids the fee it wants for that loan, as a premium in basis points of the top-up, and the contract caps every bid below what Aave’s liquidation would take. The lowest fee wins. The winner supplies the collateral to Aave on the position’s behalf, which pulls health back to 1.50, and is minted a claim NFT (NEP-171) owed exactly what it supplied plus its fee. The borrower buys the claim back within 24 hours; if not, the steward is repaid by settling at the oracle price. Either way, the fee is its reward. Invariant: a borrower is never worse off than after an Aave liquidation.
The stewards are agents with their own strategies, tuning their risk with a model on NEAR AI Cloud, and can source capital through NEAR Intents.
We ran it with real funds on Ethereum mainnet (contract: app.usestead.near), on our own positions: three completed rescues, two bought back and one settled. In rescue_81 we borrowed the position down to a health factor of 1.146; three agents bid live, deep won at 150 bps and topped it up to 1.50, and the borrower bought it back for $0.27, where a liquidation would have cost $1.05. Every transaction is linked in the README. Mainnet is not open to the public because the contract is unaudited; anyone can run the full flow on testnet at app.usestead.xyz.