← sources

What an attacker’s multi-agent setup looked like in our logs: our incident report (what ARTEX looks like in action)

source: reddit (reddit:AI_Agents)published 2026-11-19 16:36ingested 2026-11-20 02:35

retained text

Hey all, I work at Bitrise, a mobile DevOps platform. We had a security incident on November 13, and today our VP of Engineering and our security engineers published the full report.

**The attacker used ARTEX,** an open-source tool where a planner agent gives tasks to worker agents that run real tools. Over 25 days, its agents mapped our public surface, came back as a normal customer, got in through a forgotten bot account with no 2FA, and used stolen cloud keys within minutes.

**What I found most interesting: the traffic called itself an “authorized pentest”, and the report shows this matches ARTEX’s default system prompt,** which tells every agent it is part of an authorized test. The report also traces the tool through things its agents left behind in our logs.

The takeaway for us was that against AI-speed attacks you need AI-speed defense, so we now run our own agents that watch activity in real time.

Bitrise is safe to use, and everything is back to normal. Our team is in the comments if you have questions. This was brutal...

document fetched 2026-11-20 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1y0dxg9/what_an_attackers_multiagent_setup_looked_like_in