Show HN: Generate malicious CloudTrail logs with AI agents

Hi HN,

I used AI agents to simulate malicious and normal CloudTrail logs. Was playing around with automated Cloud pentesting tools (stratus-red-team) to generate attacker behavior, but there wasn't a good way to generate "noise" AKA false positives. I needed both good and bad CloudTrail logs to fine-tune my AWS anomaly detection model.

Decided to throw together a few AI agents to generate it for me. You can play around with the attack builder here: https://simulation.tracecat.com

Anybody here try using LLM agents to generate synthetic data for security analysis?

If there's interest, happy to open source the prompts and code.