YieldGarden

TL;DR — A staking protocol where users lock SEED tokens and earn ETH rewards that accrue every second. Reward math is a single timestamp per user, so every operation is constant-gas regardless of how many stakers exist. Deployed to Sepolia, 20 passing Foundry tests.

Problem

Naive staking contracts recompute each user’s rewards from full history or iterate over stakers — cost grows with users and elapsed time until the contract is too expensive to use.

Approach

Each stake stores just two things: the amount and the stakedAt timestamp. Pending rewards are a pure function of stake × rate × elapsed time, so there is nothing to loop over and nothing to accumulate globally. Claiming resets stakedAt to now — the checkpoint — so rewards start accruing fresh from that moment.

Security

Reward-bearing paths follow checks-effects-interactions: claimRewards moves the stakedAt checkpoint before sending ETH, and unstake clears the stake and decrements totalStaked before any transfer, so a re-entrant call finds nothing left to claim. Custom errors gate the edge cases (no stake, no rewards, empty reward pool).

function pendingRewards(address user) public view returns (uint256) {
    StakeInfo memory info = stakes[user];
    if (info.amount == 0) return 0;
    uint256 elapsed = block.timestamp - info.stakedAt;
    return (info.amount * rewardRatePerTokenPerSecond * elapsed) / 1e18;
}

In claimRewards, the checkpoint is advanced before the external call:

stakes[msg.sender].stakedAt = block.timestamp; // effects before interaction
(bool ok,) = msg.sender.call{value: reward}("");
if (!ok) revert TransferFailed();