Oracle-manipulation-proof borrow cap
A economic risk factor in the v1.7.0 rubric. Measured per protocol on a s cadence.
Methodology how we score #
**What this measures** This factor evaluates whether the protocol enforces a per-asset borrow cap that is calibrated to the manipulation-resistance of the oracle for that asset. Specifically, it checks whether the maximum single-transaction borrow amount is less than or equal to the oracle source pool depth multiplied by a manipulation-resistance multiplier (typically set at ten to twenty times). Sources include on-chain reads of borrow caps and DEX subgraph data for oracle source pool liquidity. This factor applies to lending protocols only.
**Why it matters** Oracle manipulation attacks are profitable only when the borrowable value exceeds the cost of manipulation. Mango Markets ($115M, 2022) is the most extreme example in the dataset: the attacker used $5M in USDC to pump the MNGO token price thirty times higher, then borrowed against the inflated unrealized profit as collateral, draining all borrow liquidity across the protocol. An oracle-manipulation-proof borrow cap would have limited the attacker's maximum extractable value to a fraction of the oracle pool depth, making the attack economically non-viable. Without per-asset borrow caps, lending protocols with thin oracle liquidity are exposed to unbounded loss from oracle manipulation.
**Green / Yellow / Red** Green: all assets have borrow caps set at or below the oracle pool depth times ten, with caps enforced on-chain and updated when oracle pool depth changes materially. Yellow: borrow caps exist but are set above the oracle pool depth times ten for one or more assets, or caps are not updated when oracle pool depth changes. Red: no per-asset borrow caps, or borrow caps are set at protocol-wide maximums unrelated to oracle pool depth.
**Common gray cases** For assets with deep, multi-venue liquidity (e.g., ETH, USDC), the manipulation-resistance threshold is effectively unlimited and this factor scores green by default. The critical scoring cases are governance tokens and long-tail assets used as collateral.
**Notable historical examples** - **Mango Markets** ($115M, 2022): No oracle-manipulation-proof borrow cap; attacker borrowed all available liquidity against a self-inflated MNGO position.
Measurement what to look for #
Determine whether the per-asset borrow cap is ≤ (oracle pool depth × manipulation-resistance multiplier).