We let an AI agent execute Bash and lived to talk about it
SessionEngineering trackconfirmed
We let an AI agent execute Bash and lived to talk about it
- Day
- Day 4 — Session Day 3
- Time
- 2:25pm-2:45pm
- Room
- Main Stage
- Track
- Harness Engineering
Accessible with the Engineering pass and above.
About this session
PostHog's Wizard agent can read your codebase, install packages, and run shell commands on your laptop. Yes, on purpose. This talk covers how we went from "defense-in-hope" to a standalone, robust security service. It'll highlight results from a pentest that made us question our life choices, an internal audit that challenged our architecture, and the debate over how to secure the entire pipeline. You'll learn why "scan-then-trust" is a weaker model than you think, what it takes to build kill switches you hope you never use, and what happens when you pentest an AI agent that has access to Bash.
Topics
Evals & ObservabilitySecurity
