Sometimes the easiest way past a system’s defenses isn’t a exploit — it’s just asking nicely, as the right person.

This is a write-up of my thought process when solving the ‘The Concierge Knows Too Much’ challenge on TryHackMe.

How a single agent tool call can steal cloud credentials, and why egress enforcement is the only reliable place to stop it.

Before an AI agent reaches customer records, cloud resources, ticketing, or internal workflows, leadership needs evidence that identity…

David Deutsch, AI and Offensive Security

Introduction

Difficulty: Easy Category: AI/LLM security

She knows your name, your room, your coffee order, none of which you told her. Word your next question carefully and she’ll also hand over…

How we run AI-generated modules beside patient data without trusting every line

Everyone’s hardening LLM agents against prompt injection. The bug that actually leaks your data is broken object-level authorization…