QuillAudits · Multisig inspector
Real multisig incidents and how independent decoding, hash verification, and Safe configuration scanning could have prevented them. Scroll to watch the counter tick down.
Attack vector
3-of-6 and 3-of-5 keys were all on one laptop; attacker built an offline Safe transaction for a malicious bridge upgrade.
Would the tool have helped?
Tool can't detect key co-location, but transaction/implementation drift monitoring would surface the malicious bridge upgrade and implementation change — with alerting, possibly in time to react.
Attack vector
Minting multisig had a 1-of-3 threshold; one compromised key gave full control.
Would the tool have helped?
Clean catch. This is exactly what Safe Inspector flags: 'threshold of 1 removes multisig protection' is a stock High finding. A single scan at any point pre-hack would have surfaced it.
Attack vector
Attacker modified the airdrop destination address to a lookalike (0x70AE678b457C5E1b3fD7AD9537F234dFc1795C15 instead of 0x70ae7D3DECfB4C3aE996fb1c07092566F73D5c15).
Would the tool have helped?
Clean catch. Transaction Inspector reveals the exact destination address in the decoded calldata, exposing lookalike substitutions before a signer approves.
Attack vector
Signers were socially engineered into pre-signing hidden authorizations; zero-timelock council migration.
Would the tool have helped?
Solana / Drift is out of scope today. The underlying issue was socially engineered pre-signing, which no on-chain tool can prevent once signers approve.
Attack vector
Single insufficiently protected key enabled unauthorized minting at scale.
Would the tool have helped?
Not a signing-review failure, but a Safe-Inspector-style config scan would flag single-key control over minting (equivalent of a threshold-1 finding) before the hack, if anyone ran it.
Attack vector
Executive device compromise; keys drained from a Solana multisig.
Would the tool have helped?
Solana / Squads is out of scope today. This case supports expanding chain support in future.
Attack vector
Leaked admin key led to a malicious ProxyAdmin upgrade.
Would the tool have helped?
Not a Safe-signing failure, but implementation/upgrade monitoring of the pattern this tool checks would have flagged the ProxyAdmin change instantly.
Attack vector
Retained developer admin privileges were abused.
Would the tool have helped?
A Safe-Inspector-style privilege review ('is this owner/role expected?') is precisely the check that would surface a leftover privileged address.
Attack vector
Compromised Safe{Wallet} UI; signers blind-signed a delegatecall that swapped the Safe's masterCopy.
Would the tool have helped?
Flagship case. Independent decode would have shown a delegatecall + implementation change instead of a routine transfer; recomputing the SafeTxHash on a clean device would not match the intended transaction. Exactly the gap this tool targets.
Attack vector
Compromised signer devices showed a benign transaction while the actual payload transferred pool ownership.
Would the tool have helped?
Decoding on a clean second device would reveal the ownership-transfer call. Caveat: the devices themselves were compromised, so verification had to happen off those machines.
Total left
$2.06B
