Brilliant. It's so cool that you and Keshav were truely ahead of the curve already building something similar to these new Anthropic skills! I'm super curious about the mechanism for how they can be weaponized; is it mostly through malicious scripts, or can the metadata alone pose a risk for context injection?
Brilliant. It's so cool that you and Keshav were truely ahead of the curve already building something similar to these new Anthropic skills! I'm super curious about the mechanism for how they can be weaponized; is it mostly through malicious scripts, or can the metadata alone pose a risk for context injection?