// blog
Writing
-
Prompt injection in agentic AI tooling
When an LLM can call tools, untrusted text becomes untrusted instructions. Treat the model like a confused deputy.
-
Buffer overflows didn't die, they moved to IoT
The memory-safety bugs everyone declared dead are alive and well in embedded firmware. Here's why.
-
Mapping attack surface with passive recon
How I build a picture of a target before touching it — domains, certificates, and the trails organisations leave in public.
No posts match those filters.