Loading...
Loading...
This morning a Pendle PT lending market was manipulated. A push of less than 3% in the pool price liquidated twelve borrowers holding roughly $36M in positions. LlamaGuard PT, which we're bringing to Aave, prevents this attack vector. Borrowers in the PT-reUSD lost about $1.35M between them. Every one of those positions was healthy minutes before. The tightest buffer was half a percent, the widest about two. The design accepts a tradeoff, lenders are insulated while borrowers bear the risk. The affected markets price PT collateral from a short TWAP of the Pendle pool, taking the lower of that TWAP and a fixed linear discount. But it also connects borrower health factors directly to pool conditions. Whoever can move the pool for a single TWAP window moves the oracle, and in a PT looping market the whole book stacks near the LTV ceiling by design. That's why a sub-3% move made essentially every leveraged position liquidatable at the same moment. It's a defensive market, and the defense is paid for in user experience: holding a position safely means keeping a wide buffer and watching the pool around the clock. The Aave market, powered by LlamaGuard, is defensive. The quoted price is deliberately conservative. The difference is in how the oracle is allowed to move. It never quotes the pool directly. It quotes a linear discount that re-rates toward a smoothed, multi-day average of the market rate, and only when the drift crosses a published 0.30% gate, no more often than a minimum delay allows, in steps bounded by limits that live in contracts Aave Governance owns. A borrower's health factor moves gradually and predictably. Run today's attack against LlamaGuard's design. A fifteen-minute pool push barely dents a multi-day smoothed rate. Even a sustained displacement buys the attacker one bounded step per delay window, each with limited price impact. Walking the oracle down far enough to force a cascade means holding the Pendle pool displaced for days, in public, while every step prints on-chain and the risk stewards and guardian watch it happen. The attack doesn't become impossible. It becomes slow, expensive, and visible, with the payoff throttled to a trickle. Manipulation is an economics game, and the economics invert. A second layer compounds this. On the Aave market the liquidation threshold is dynamic: it rises toward its cap as the PT approaches maturity while the liquidation penalty falls toward its floor. Open a position near the line and, all else equal, your health tends to improve with time, both because the PT pulls to par and because the threshold walks up beneath you. Under a static configuration, a tight position only ever decays. Put together: the protocol is protected by the dynamic liquidation threshold, borrowers are protected by capped and predictable oracle steps, and the market's stability doesn't hang on a pool that can turn illiquid or volatile without warning. That's what a risk layer is for. The full argument, and the architecture behind it, is in the piece below. 👇
Source:https://x.com/StaniKulechov/status/2092554777129357651
Impact Score