Parameter Recomendations
Posting parameter recommendations as well as a short note from the RWG on the risk work behind this one.
| Supply Ceiling | 5,000,000 |
| Daily Borrow Limit | 500,000 |
| Collateral Factor | 87% |
| Liquidation Factor | 100% |
| Liquidation Incentive | 5% |
| Minimum Debt Amount | 3000 |
This reintroduction is the output of a full risk refresh, not a parameter tweak. We held crvUSD-LP collaterals offboarded since February specifically until the Yield Basis migration that drove the original risk had matured, and we treated the reassessment as a real question with a chance of staying closed. What changed our view was evidence rather than time passing: the v3 pools redesign targets the exact failure mode behind the offboarding, and June’s ~23% BTC drawdown let us measure it live rather than in simulation. The temporary redemption discount that ran to 41–45% for a month under the old design peaked near 4.8% and cleared in about three days under v3, at comparable drawdown depth.
The parameters here are deliberately conservative, an 87% CF and a 5% liquidation incentive, sized so that a worst-case redemption discount at the depths v3 hasn’t yet faced is absorbed without loss to FiRM. Periodic reviews and a phased approach lets exposure grow as lindy accrues without forcing borrowers to deleverage if our read changes. We’d rather under-size and earn the deeper observation on live exposure than overextend and have to backtrack.
This is also FiRM’s first market on a Stake DAO OnlyBoost v2 escrow, so it carried its own due-diligence track alongside the crvUSD work. The escrow stakes the LP into a RewardVault that we verified on-chain as a strict 1:1, non-upgradeable wrapper, which means the collateral always unwraps to the exact LP quantity staked and yield is paid as separate reward tokens rather than compounded into the price, leaving no share-price manipulation surface on the valuation. We reviewed the four public Stake DAO audits against FiRM’s liquidation model and found no Critical or High residual to FiRM; the two Pashov Criticals target a Stake DAO lending product FiRM doesn’t use. The residual risk concentrates in one place, the integrity of the Stake DAO withdraw path, since the 1:1-no-haircut design means a Stake-DAO-side shortfall surfaces as a liquidation revert rather than a graceful markdown. Most of that is shared with our existing Convex escrows; the net-new piece is a custody-substitution lever (re-pointing the staking strategy) that sits behind a delay window governance Timelock fronted by a single multisig. We treat that 5-day window as the defense, not the multisig, and we’ve built a dedicated monitor on the shared Stake DAO Timelock that pages on any unexpected custody op so we can trigger the escrow’s guardian withdrawal, pulling positions to plain LP, before a hostile change could mature. We’ve also raised two hardening asks with the Stake DAO team directly: making the Timelock its own admin, and granting cancel rights to an independent security council so the delay becomes a true veto rather than just an observation window.
The full reasoning, contract-surface analysis, Yield Basis counterparty and v3 work, the live June stress evidence, the oracle and Stake DAO escrow design, and the parameter rationale, is in the reassessment document. Happy to take questions here.