Compromised Flowise became a staging point for adaptive scripts aimed at other servers

Wiz inspected a compromised Flowise deployment and reconstructed approximately ninety-nine generated scripts within twenty-four hours. The path began at a custom-function endpoint with an internal-request header, then reached command execution through the host’s available runtime. Subsequent scripts probed other Flowise instances and experimented with Node inspector and process-signal paths when earlier attempts failed. This is a documented compromised host and sequence of attempts, not evidence that all scanned hosts were compromised.

The use of adaptive, repetitive scripts suggests automation with feedback, but the source does not supply a cryptographically proven model identity. The useful boundary is the service accepting executable functionality and granting the resulting process access to additional network targets. Defending that path requires checking authorization on the actual execution route and constraining what a compromised runtime can reach, rather than assuming generated-script similarity alone identifies an attacker.