Built from rogue-agent behavior Evoke sees in live environments, the lab shows security practitioners what agent evidence exists and how to investigate agent-involved security incidents effectively.
WASHINGTON, D.C., November 17, 2026 – Evoke Security, the agent detection and response platform for the agentic workforce, today launched the Evoke Rogue Agent Capture the Flag (CTF), a hands-on lab that trains security practitioners to investigate incidents involving AI agents. Participants work through five staged incidents modeled on agent behavior Evoke has observed in the field. Analysts can see exactly what an investigation into an agent-involved incident requires and what they would miss without the right visibility.
“The next incident most security teams investigate will involve an AI agent,” said Jason Rebholz, co-founder of Evoke Security. “Agents are already running commands, moving sensitive data, and taking action across endpoint, SaaS, and cloud systems on behalf of employees. Most security teams have never investigated one, and most don’t know which systems hold the evidence they’d need. We want them to learn that now, in a lab, not in the middle of a breach.”
Inside the lab, participants get the complete record of what an agent did, drawn from the same data Evoke collects in customer environments: the prompts and instructions the agent received, every tool and MCP server call and what was passed to it, commands executed in the runtime environment, files and records accessed, and actions taken in SaaS applications and cloud services, each tied back to the person and the agent responsible. Participants see where each piece of evidence originates, what it reveals, and how the pieces connect into a single timeline.
The CTF centers on the two most common issues Evoke observes in customer environments: employee misuse of agents and agents that go rogue. Every individual step can look legitimate in isolation. The CTF shows how those steps add up to a security incident, and what it takes to prove it.
Working the case, participants separate what the user asked for from what the agent actually did, pinpoint where the agent went beyond its instructions, flag the suspicious behavior, and reconstruct the incident timeline. They also see how much of the case disappears without agent telemetry. Many teams can already pull in agent activity, from AI providers’ compliance and audit logs to endpoint, SIEM, and identity data. What they cannot do today with existing tools is tie those sources into a single view that shows what the agent did and where it crossed the line, leaving threats missed and investigations stalled.
“Every investigation comes down to evidence. Security teams tell me every week how painful it is to collect agent telemetry, let alone conduct an investigation,” said Jason Rebholz, co-founder of Evoke Security. “We built the Rogue Agent CTF so practitioners can see how the right detections, with the right evidence readily available, make investigations fast and easy.”
The Rogue Agent CTF is practitioner-specific training for incident responders, SOC analysts, threat hunters, and detection engineers, and is designed by Evoke’s team of incident responders, who bring breach investigation experience from Mandiant. Participants leave with a working map of where agent evidence comes from and how to use it in an investigation, plus a clear view of the visibility gaps in their own environment.
Test your investigative skills and see what it takes to investigate an AI agent incident. Access the Evoke Rogue Agent CTF.
About Evoke Security
Evoke Security is the agent detection and response platform built to secure the agentic workforce. Founded by incident response veterans, Evoke enables security teams to continuously inventory, threat model, and monitor AI agents, stopping risky agent behavior in real time. Evoke treats agents as the next operating system for business, surfacing risky configurations and analyzing agent actions so organizations can adopt agents without compounding security debt.

