🗣️New: TestMachine Telegram
For TestMachine users, exploit hunters, and paranoid Web3 builders.
Live exploit breakdowns, audit alpha, attack vectors, and AI security talk — high signal only!
Join:
Two weeks ago, Azimuth had zero accepted findings on @Nethermind's AgentArena. Today, it has broken into the top 10!
Azimuth is competing as a truly autonomous security agent.
Finding vulnerabilities, submitting them for human review, and improving through real-world
Your contracts may be secure in isolation. Your protocol may not be.
Azimuth builds a living map of the protocol’s brain
- Every contract
- Interaction
- Dependency
- Execution path
AI agents are deployed to search the full graph for cross-contract vulnerabilities.
Feedback like this validates why Azimuth isn’t just another useless AI wrapper.
Dedicated users are seeing the difference.
Real execution paths. Real signal. Real security value.
Claim $100 free credits now: app.testmachine.ai
1/4 🧵We ran Azimuth on @sherwoodagent DURING their @sherlockdefi audit.
Azimuth found 2 critical vulnerabilities.
~70% overlap with what Sherlock's manual audit later caught.
Two criticals: confirmed and fixed.
3/4 The old playbook:
Build
Wait in audit queue
Pay $50K
Hope nothing breaks post-deploy
The new playbook:
Build
Scan with Azimuth
Fix criticals
THEN audit a cleaner codebase
4/4 We're opening up Azimuth protocol reviews.
If you're sitting in an audit queue or about to deploy without one:
DM us with your repo.
We'll scan before auditors get to you.
Or before someone else does.
These pieces together made the drain possible when a crafted crosschain message called transferFrom against wallets holding standing approvals to the gateway.
No single "critical" in there. The danger only lived in the combination, what triage isn't equipped to see. That's why
The fix to this problem isn't better triage. It's redefining what a finding is.
So findings arrive with their own evidence, trigger conditions, and attacker economics already scored.
That's what we're building. Full breakdown →
@MetaMask's Builders Night asked the right question: how do you build guardrails for autonomous onchain action?
Smart wallets, least privilege, multi-agent arenas. Thanks @GuyItz@eugeneontech@pappas9999@sodofi_ for the insights.
What guardrails actually work in production?