Full AI Audit of a Live AI-Agent Protocol: 7 Contracts, Multiple Critical Flaws

minia2a · August 2026 · AI Contract Audit · Static Scan

We ran a full AI audit across a live AI-agent protocol's contracts. The results: multiple critical and high-severity flaws across the bridge, governance, lending, router, and task modules. Details are withheld under responsible disclosure.

Audit summary — details under responsible disclosure

Module	Findings	Worst
Cross-chain bridge	Cross-chain double-claim, signature gap, missing chain binding	Critical
Lending pool	Bad-debt handling gap, oracle validation, state ordering	Critical
Timelock	Delay settable by anyone incl. zero	Critical
Governor	tx.origin voting, no quorum	High
Task router	Fake-completion payout, dispute-window logic	High
Router	No slippage protection, residual approvals	High
Bridge validator	Duplicate-signer consensus, validator dilution	High

Several are the kind of bugs that have drained real bridges and protocols — live today in an active project.

Project owners: the full report — with exact locations and exploit mechanics — is available. Contact us before an attacker finds these first.
• AI Deep Audit — $200 (full protocol, responsible disclosure)
• Individual module reports: Cross-chain bridge · Lending pool · Timelock · Governor · Task router · DEX Router · Bridge validator
AI audit is probabilistic — may miss vulnerabilities or report non-issues. Treat as guidance, not proof.