TARE
Multi-collateral CDP with redundant oracle checks, bad-debt Dutch auctions, sTARE savings, Curve PegKeeper, and surplus routing into the reinvestment loop.
TARE case study →Live at Sepolia/testnet scale and internally composition-reviewed. Not externally audited. No mainnet TVL. Routers move value already released or earned — they do not mint from nothing.
Multi-collateral CDP with redundant oracle checks, bad-debt Dutch auctions, sTARE savings, Curve PegKeeper, and surplus routing into the reinvestment loop.
TARE case study →USDC multi-strategy ERC-4626 vault with senior/junior tranches. FLYWHEEL 2.0: receives Coil keep_bps fees + TARE surplus emission slice. CoilMakerStrategy is PRIMARY.
Keep case study →Vote-escrow tokenomics, gauge weights, bribe distribution, and GaugeWeightRouter steering Keep allocations via set_allocations().
Intent-based batch-auction DEX. EIP-712/EIP-1271 order flow, partial fills, replay protection. Trading fees enter CoilFeeRouter for the 3-way split.
Coil case study →Permissionless routers sit at the seams. They hold no admin keys to user collateral — only route fees and surplus already earned.
Batch settlement accrues USDC protocol fees.
keep_bps (0–50%) → Keep vault · gauge_bps (0–50%) → BribeDistributor · remainder → Curve → TARE surplus.
Senior slice (≥70%) → sTARE yield · emission slice (≤30%) → Keep vault reinvestment.
veTARE votes steer capital across strategies within each strategy's max_debt.
Primary path: EIP-1271 market making deepens books → more volume → more fees.
No vote, keeper, admin, or caller may move a solvent user's collateral, mint on a bad price, or alter live debt. That claim is engine-scoped — Keep depositor funds still depend on admin custody, strategy limits, and solver honesty.
Composition Auditor, Deploy Verifier, Watchtower, DiffRisk, StressSim, Overseer, and Herald (human-gated). Read-only on the critical path.
Agent suite →Inventory-funded Aave V3 liquidation executor + bot kit. Outside the flywheel value path — explicit owner-key and inventory risks.
SnekSentry case study →700+ tests across repos · Hypothesis/Titanoboa stateful fuzz · fork-proven on Sepolia · self-reviewed, not audited.