Moonwell Freezes Base Borrowing After $8.7 Million MAMO Price Manipulation Exploit
27 disclosure and independent tracing from PeckShield and CertiK, an attacker inflated the reported value of MAMO on the lending protocol's Base deployment to borrow approximately $8.7 million in cbBTC before consolidating proceeds into DAI. Borrow caps across all Base Core Markets have since been reduced to 1 wei as a precautionary circuit breaker, effectively suspending new borrowing activity while the investigation continues.
The transaction lifecycle
The attack decomposed into four discrete state transitions, each exploiting the structural assumption that MAMO's on-chain price could be trusted as a collateral reference. Blockaid, monitoring the transactions in real time, identified the same mechanism independently and initially reported a drainage of roughly 50.6 cbBTC — valued at over $4 million at the moment of capture — before PeckShield's subsequent tally revised total losses upward to approximately $8.7 million across the complete transaction graph.
In the first phase, the attacker accumulated a position in the thinly traded MAMO market and executed localized trades sufficient to move the price reference fed into Moonwell's pricing logic. In the second phase, the inflated collateral valuation propagated through the protocol's market configuration, raising the effective borrowing capacity attached to the attacker's MAMO deposit. The third phase consisted of borrowing real cbBTC against the artificially valued collateral from the mCBTC market — assets drawn from a deep-liquidity pool against collateral whose reference price had been locally distorted. In the fourth phase, the attacker converted the borrowed cbBTC into DAI and consolidated the proceeds at a single address, terminating exposure.
The systemic vulnerability is not novel: oracle price feeds that derive from a single DEX source or a narrow trading window inherit the manipulability of that source. When the referenced market lacks depth, a bounded amount of capital is sufficient to move the reported value by orders of magnitude larger than the cost of the manipulation itself. For lending protocols operating in adversarial environments, this asymmetry defines the upper bound of safe collateral integration.
Prior failures and configuration drift
Thursday's incident is not the first oracle-related failure on Moonwell's Base deployment. In February 2026, a pricing failure across the protocol's lending markets left approximately $1.78 million in bad debt after an oracle calculation error mispriced Coinbase Wrapped ETH (cbETH) at roughly $1.12 while the asset traded near $2,200. The distorted valuation enabled liquidators and automated bots to repay positions at the corrupted price and seize cbETH collateral at a fraction of its market value. The faulty oracle logic, per the protocol's disclosure, included code generated with Anthropic's Claude Opus 4.6 model, and the error was traced to an incorrect scaling factor in the calculation path.
The recurrence pattern is worth noting. Both failures concentrated at the oracle boundary — the interface at which off-chain price reality is mapped onto on-chain collateral valuations. In the cbETH incident, the failure was computational: a scaling factor produced a price several orders of magnitude below spot. In the MAMO incident, the failure is informational: a manipulable price source was treated as a trustworthy reference. Different mechanism, same architectural seam.
The protocol's response on Thursday also bears observation. Borrow caps across all Base Core Markets were reduced to 1 wei, not merely the MAMO market where the issue was identified. Supply caps for MAMO and WELL were similarly reduced to 1 wei, while supply limits for other assets were left unchanged. This is a coarse-grained containment strategy: it stops new borrowing across the deployment entirely rather than isolating the specific market where the manipulation was executed. The trade-off is operational availability versus blast radius, and the protocol has prioritized the latter.
What the oracle layer exposes
For developers building lending primitives in the Base ecosystem, the incident surfaces a recurring question about collateral onboarding criteria. MAMO's history provides a clear retrospective signal: the token fell sharply after its Coinbase debut in August 2025, having gained more than 120% in the preceding week before reaching an all-time high of $0.227 and then losing nearly 20% as selling activity intensified. High pre-listing volatility and concentrated post-listing drawdowns are themselves risk indicators that should be priced into collateral eligibility decisions, not merely observed after the fact.
The broader lesson concerns the oracle integration surface itself. Price feeds that sample from a single source, or that aggregate across sources with insufficient volume weighting, replicate the manipulability of their weakest input. For protocols that depend on continuous liquidation viability, the price reference is not a convenience but a liveness-critical dependency; degrading its reliability degrades the protocol's ability to honor obligations under stress.
Whether Moonwell's Base deployment returns to its previous configuration or migrates to a more hardened oracle stack — with deeper liquidity sampling, time-weighted average pricing, or multi-source aggregation — will determine the protocol's next-state viability on this chain. Two oracle-related failures within six months indicate that the current configuration does not hold under adversarial pressure, and that the boundary between computation and information at the oracle seam remains the weakest link in the architecture.