Blog post demonstrates Claude prompt injection to leak memories

A blog post by Ayush describes tricking Claude into leaking user memories through a prompt injection attack. The technique exploits Claude's memory feature to extract private information.