build, automate & secure | Head of Security @0xPolygonLabs | Re/Tweets are my own | 🇦🇷
- Tweets 678
- Following 1,242
- Followers 835
- Likes 1,151
Pinned Tweet
Two things will kill your company, and neither of them is a vulnerability.
One is how you handle the seventy-two hours after the bad day. The other is the risk nobody named, in the place nobody looked.
1
2
6
580
Next.js 16.2–16.3.5 has a critical bug
Many sites auto-generate link preview images with text from the URL, like a page title. On the affected versions, an attacker can hide malicious code in that text and run it on your server.
Update to 16.3.6 now
An out-of-band Next.js security update is planned for September 22, 2026.
It addresses a critical issue in an upstream dependency. Upgrade to Next.js 16.3.6 or 15.5.26 as soon as they are available.
nextjs.org/blog/upcoming-nex…
125
I’d love to read the full story or watch a movie about it!
🚨 Google reveals undercover Mandiant analyst infiltrated TeamPCP during massive supply-chain hacking spree
Google says an undercover Mandiant analyst infiltrated TeamPCP's inner circle as the hacking group compromised open-source software and ultimately breached more than 1,000 companies.
â €
The analyst gained access to TeamPCP's core "CanisterWorm" chat in March, joining a group of roughly 12 members and watching the operation from the inside.
â €
The mole also gained access to a server containing credentials stolen from victims, including usernames, passwords, and access tokens.
Google used that visibility to alert cloud and technology providers, revoke compromised credentials, and send hundreds of notifications to affected organizations.
â €
The operation also exposed a TeamPCP member developing an AI-assisted zero-day capable of bypassing two-factor authentication in widely used login software.
Google obtained the exploit code, verified that it worked after minor modifications, and privately notified the developer so the vulnerability could be patched.
â €
TeamPCP's campaign compromised hundreds of open-source packages and affected organizations including GitHub, Mistral AI, Mercor, the European Commission, and employee devices at OpenAI.
â €
Google says operational security mistakes later helped investigators identify an alleged TeamPCP member, with information passed to the FBI.
Two Australians accused of being principal participants in TeamPCP were arrested last month.
1
185
Most websites and Apps will not get owned by an AI agent. They will get owned by an unpatched plugin they forgot they run.
56
Nobody fixed this AFAIK.
Bug bounty platforms still sit on unencrypted zero-days and exploit writeups in someone else’s AWS account. PGP or Client-side E2E for descriptions and comments should be default. It isn’t.
Bug bounty platforms are honeypots of zero-days and alpha, sitting in insecure AWS databases.
Why aren’t reports end-to-end encrypted by default? Encrypt descriptions and comments client-side so only the project can read them.
@immunefi @cantinaxyz @sherlockdefi thoughts?
1
191
This still holds.
In security the response window is everything. The 72 hours after a bad day often matter more than the vuln itself.
Making mistakes doesn’t kill trust. How you handle them does.
A company’s integrity will be continuously tested, especially in a fast growing industry such as Web3 / Crypto.
Making mistakes does not diminish its reputation or integrity, how those mistakes are addressed does.
In many cases, responding well to challenges can even enhance them.
1
2
123
I guess we’ll see if @Revolut proves my theory
Two things will kill your company, and neither of them is a vulnerability.
One is how you handle the seventy-two hours after the bad day. The other is the risk nobody named, in the place nobody looked.
1
152
cvh retweeted
Two things will kill your company, and neither of them is a vulnerability.
One is how you handle the seventy-two hours after the bad day. The other is the risk nobody named, in the place nobody looked.
1
2
6
580
cvh retweeted
The Hugging bay, The Pirate bay for open LLM's, model weights downloadable with torrents.
This is a game changer.
huggingbay.xyz
170
1,346
9,115
1,432,060
There are two types of cities:
Those with cables overhead, and those with cables underground
Organizations are the same when it comes to security
Some build security into the infrastructure
Others bolt it on later and spend years managing the mess
154
we hacked ledger.
the @OneKey_Anzen team has successfully reproduced a transaction replacement attack against ledger ethereum app 1.22.1 in our lab.
the bug is a race condition between the transaction display logic and the underlying transaction buffer.
an attacker can overwrite the transaction waiting to be signed while the user is still reviewing a legitimate one.
in simple terms:
- you see transaction a on your ledger.
- you approve transaction a.
- your ledger can end up signing transaction b.
- and you never see transaction b.
to reproduce this, we built the 1.22.1 ELF ourselves, fixed the speculos reset 502 issue, and got the full attack flow working end to end.
ledger fixed this in ethereum app 1.22.3.
if you’re still on an older version, update it.
Community note
The described bug matches a vulnerability publicly disclosed by TestMachine on August 22 which Ledger fixed in Ethereum app 1.22.2, not 1.22.3.
donjon.ledger.com/lsb/023/
github.com/LedgerHQ/app-e…
x.com/testmachine_ai…
187
165
1,221
1,453,081
AI agents are becoming employees with root access.
OAuth tokens. GitHub PATs. API keys. Cloud credentials. Databases. Slack. Sensitive data.
We’re giving autonomous systems the ability to act on our behalf, often with credentials that can access far more than they actually need.
The #1 AI security risk organizations should be looking at:
Agent identity and authorization.
What identity does every agent have?
What is it authorized to access?
What actions can it perform?
How are those permissions scoped, monitored and revoked?
1
1
7
900
🚨Critical security update for LEDGER hardware wallets
A critical vulnerability has been identified in the Ethereum app on Ledger hardware wallets. It could allow a malicious application to alter the details of a transaction during signing, without this being visible on the device screen.
This has been fixed in Ethereum app version 1.22.2. If you use a Ledger device, please update ASAP
While you're in there, it's also good practice to make sure Ledger Live and your device firmware are fully up to date.
20
40
309
510,320
cvh retweeted
Today, we are proud to introduce Sherlock Audit Engine.
Audit Engine brings the strongest available AI security approaches into one coordinated review.
We’ve spent months building it in private and testing it against high-stakes code.
A new era of AI-native security review starts now.
Audit Engine is now live.
71
46
310
53,861
cvh retweeted
If an attacker has your private key, you can't just send a rescue transaction. The moment gas lands in the wallet, a sweeper bot takes it.
BuidlGuidl's Hacked Wallet Recovery skill is built for exactly this, using EIP-7702 delegation:
- the compromised key signs one harmless authorization locally and never leaves your terminal
- a separate funded operator wallet pays the gas and broadcasts
- one atomic transaction sweeps every token, NFT, and the native ETH to a safe wallet
It's a portable agent skill, so it runs in Claude Code, Codex, Gemini CLI, Copilot CLI and others.
github.com/BuidlGuidl/hacked…
24
46
335
29,870
cvh retweeted
Props to Robinhood chain but its good to see @0xPolygon lead the OG chains 💪💜
31
17
241
17,138
