LLMs, WAF Bypass, LLMs...

A great article, definitely worth a read. It’s also worth looking through the linked GitHub repository to learn a few tricks On the Coming Industrialisation of Exploit Generation with LLMs.

It feels like /.well-known/ isn’t that well known by WAFs… What I really like about this post is that it highlights one of the key tricks for finding vulnerabilities: “What routine tasks may open a security hole.” Cloudflare Zero-day: Accessing Any Host Globally.

I initially saw that trick on LinkedIn, but this article gets into much more detail. The magic string that "breaks" Claude Claude Magic String Denial of Service.

GitHub Security Lab explains how they built an LLM taskflow system to triage CodeQL alerts: small, repeatable tasks, stored intermediate state, MCP tools for deterministic checks, and GitHub Issues as review checkpoints AI-supported vulnerability triage with the GitHub Security Lab Taskflow Agent.