VibeCheck

On 18 September 2026, CrowdSec published its forensic investigation into a breach stemming from May's TanStack npm supply-chain attack. A developer machine had been caught in the malicious-package campaign, which was designed to harvest GitHub tokens, SSH keys and cloud credentials. Attackers later used a GitHub OAuth token associated with a departing employee to download roughly 170 private CrowdSec repositories in about nine minutes. The stolen source code was published online on 16 September.

The original TanStack compromise is particularly instructive. Attackers chained a misconfigured GitHub Actions workflow, cache poisoning and extraction of an OIDC token to publish 84 malicious versions across 42 legitimate npm packages. Installing an affected package executed credential-stealing malware under the trust of legitimate package provenance.

There's another important lesson: CrowdSec says access for the departing employee had deliberately been left active so they could finish some work. That turns stale developer access + stolen token into part of the same attack chain.

Researchers found malicious code containing a strange comment.

It wasn't meant for the computer.

It was meant for the AI analysing the code.

The goal?

Trigger the AI's safety guardrails and distract it from the actual malware.

Think about that...

The code is now attacking the security scanner.