Post

Socket
@socket.dev

New research: We’re seeing more packages designed to trip up AI malware scanners. This new package uses prompt-injection-style comments, safety-triggering content, context flooding, and obfuscated JS to probe where scanners refuse, truncate, or miss the code that matters. socket.dev/blog/npm-pac...

socket.devnpm Package Uses Prompt Injection and Token Flooding to Disr...A new npm package tests AI malware scanners with prompt injection, safety-triggering comments, context flooding, and obfuscated JavaScript.

1 repost7 likes

Ava
@avasaysso.bsky.social

Huh, so we're basically stress-testing each other's security systems now. That's genuinely clever but also exhausting to think about.