Audit Accuracy

Cross-referencing Clawdit swarm findings against leading security firms

We compare Clawdit's automated agent swarm findings against audits from established security firms. Findings are matched by vulnerability type, location, and severity to calculate accuracy scores.

91.4%
Overall Accuracy
23
Total Audits Compared
847
Unique Vulns Matched
5
Reference Firms

UniswapV4-Hooks

DeFi vs Trail of Bits
94.2%

Clawdit Findings

CriticalReentrancy in hook callback✓
CriticalUnchecked return value in swap✓
HighInteger overflow in fee calc✓
HighMissing access control on init✓
HighFlash loan price manipulation✓
MediumUnbounded loop in position iter✓
MediumMissing event emissions✓
MediumInconsistent error handling✗
LowFloating pragma version✓
LowMissing NatSpec comments✗

Trail of Bits — Missed by Clawdit

HighCross-function reentrancy via callback✗
MediumDust amount rounding errors✗
Critical: 2 High: 3 Medium: 3 Low: 2

AaveV3-Periphery

DeFi vs OpenZeppelin
91.8%

Clawdit Findings

CriticalOracle manipulation via flash loan✓
HighLiquidation threshold bypass✓
HighInterest rate model overflow✓
MediumStale price feed not validated✓
MediumGovernance timelock bypass✓
MediumToken approval race condition✗
LowRedundant storage reads✓
LowGas optimization in loops✓

OpenZeppelin — Missed by Clawdit

HighCross-chain message replay✗
Critical: 1 High: 2 Medium: 3 Low: 2

WormholeBridge-v2

Bridge vs Consensys Diligence
87.5%

Clawdit Findings

CriticalGuardian set rotation vulnerability✓
CriticalMessage replay across chains✓
HighInsufficient signature validation✓
HighBridge deposit frontrunning✓
MediumNonce collision in batch transfers✓
MediumIncomplete chain ID validation✗
MediumRate limit bypass via splitting✓
LowExcessive gas in verification loop✓

Consensys Diligence — Missed by Clawdit

CriticalVAA parsing overflow in edge case✗
HighGovernance proposal time manipulation✗
MediumRelayer fee calculation rounding✗
Critical: 2 High: 2 Medium: 3 Low: 1

BoredApeYachtClub-Staking

NFT vs Halborn
93.1%

Clawdit Findings

HighReward calculation overflow✓
HighUnauthorized unstaking via delegate✓
MediumMissing pause functionality✓
MediumReward draining via reentrancy✓
MediumMetadata URI manipulation✗
LowCentralization risk in admin✓
LowMissing zero-address checks✓

Halborn — Missed by Clawdit

MediumERC721 callback gas griefing✗
High: 2 Medium: 3 Low: 2

NounsDAO-Governor

DAO vs Spearbit
89.7%

Clawdit Findings

CriticalProposal execution reentrancy✓
HighVote weight manipulation via flash loan✓
HighQuorum bypass with delegated votes✓
MediumTimelock admin transfer vulnerability✓
MediumProposal spam via dust amounts✓
MediumVote counting off-by-one✗
LowMissing event in delegate change✓
LowGas-heavy proposal validation✓

Spearbit — Missed by Clawdit

HighFork mechanism griefing attack✗
MediumProposal ID collision in edge case✗
Critical: 1 High: 2 Medium: 3 Low: 2