Claude RedTeam, Claude Hacking, Claude Skills...Is it Claude week?

The latest Claude models are getting noticeably better at hacking AI models are showing a greater ability to find and exploit vulnerabilities on realistic cyber ranges.

Getting Claude Code to execute commands even when it's not supposed to. A great read even if you don't care about Claude Code, as it highlights a few tricks and quirks of some of your favorite CLI tools Pwning Claude Code in 8 Different Ways.

A really interesting read on the state of OpenSSL The State of OpenSSL for pyca/cryptography.

This week, Trail of Bits released a set of skills for Claude. A lot of great skills that can be used by code reviewers, smart contract auditors, and everyone working in AppSec Trail of Bits Skills Marketplace.

The team at Block wrote a proposal on how something similar to CORS can be used to prevent LLM content injection Agent Guardrails and Controls: Applying the CORS Model to Agents.

Laurent Gaffié (the author of Responder) released a new fuzzer. If you know about his previous work on SMB, you know that it's worth checking out OffByWon Fuzzing Framework.