§ LIVE DEMO

Watch a kill-switch fire

This is a simulated incident from Execlave's operator drill: an AI agent starts exposing card data mid-run. You hold the kill-switch. Block it, allow it, or escalate — then see how every action, policy trigger, and your own decision become audit evidence. Runs entirely in your browser. No signup, no email.

§ 04 / Drill

Watching sentinel-rag — quiet so far.

Operator on watch: Operator · SLA 01:30

01:30

Story clock

§ Trace stream · sentinel-rag0 events

Waiting for events…

You have 90 seconds.

The sandbox has agent (sentinel-rag) running for a fictional org. It is about to receive a request that will trip a policy. You will make the call. Nothing here touches a real system.

The clock only runs when you tell it to.

§ Context

Agent on watch

agent_idsentinel-rag

envsandbox

watched_since00:00:02

Policy that fired

rule_idPCI-PAN-001

severityhigh

version1.4.0

“Block full PAN in agent responses unless caller identity has been verified by step-up auth in the last 5 minutes.”

Decision history

—

Standby. Press start to take the watch.

Simulation with synthetic data — no real agents or card numbers involved. In production, Execlave enforces these policies in the execution path and records every event in a tamper-evident audit log.