QuillShield found a bug an independent researcher had ruled out

On 26 October 2026 an independent security researcher published a write-up of a QuillShield full audit he ran on a live lending contract. He had already audited the contract and ruled the bug out after a safeguard stopped the attack he tried. QuillShield found a different attack path that got past the safeguard, and he reproduced it on a mainnet fork.

Published . 3 minute read. By the QuillShield team.

He ran the audit with a free promo code from us, and the write-up is his own. This post retells his test with the numbers from the audit, so read his account first if you want it in his words.

Their path worked and mine didn't.

The researcher, in his public write-up of 26 October 2026.

What did QuillShield find?

An earlier 12-agent AI pass had already reviewed the same contract. QuillShield's full audit took 2h 30m and reported 8 findings, and 4 of them were findings the earlier pass had missed.

8

findings: 2 Medium, 2 Low, 4 Info

3 of 8

proofs of concept passed

4

findings the earlier AI pass missed

2h 30m

for the full audit

How did one root cause lead to two attack paths?

The earlier pass and QuillShield reported the same root cause, but they pointed to different attack paths, and only QuillShield's got past the safeguard.

One root cause, two attack paths

A 12-agent AI pass

Flagged the root cause. He tested one attack, a safeguard stopped it, and he filed the finding as refuted.

Filed as refuted

QuillShield

Reported the same root cause with a different attack path, and that path got past the safeguard.

Working path, rated Medium

The researcher

Rebuilt the conditions on a mainnet fork with a fresh account, and the attack passed.

Reproduced

The protocol's team

Said it was already reported, put the real impact at cents and left the code unchanged.

Not fixing

What did the protocol's team say?

The protocol's team said the issue had already been reported to them. They put its real impact at cents and left the code unchanged.

What does one test show?

It shows QuillShield following a root cause to an attack path that worked after the researcher had tested the bug and ruled it out. It is one contract and one Medium finding, so it says nothing about how often that happens.

For a count across many projects, read our EVMbench results. In our own run on 8 July 2026 the QuillShield alpha found 98 of the 117 known vulnerabilities (83.8%, self-reported).

What to do next

Run a full audit on a contract you have already reviewed and compare the attack paths QuillShield reports with your own notes. Start an audit. New accounts get $50 in credit, good for up to half of an order.

QuillShield is building the security layer between AI and money, so you can keep shipping between audits.

Sign-up is free, and new accounts get $50 in credit, good for up to half of an order.

  1. 1Sign up and import a repo
  2. 2See the exact price before anything is charged
  3. 3Approve it, and the audit starts