Ulk0r0
Author ProfileClaim This Author Profile
Prove ownership by publishing #HashtagPLUS and this profile link on your author page or an article under your byline. A moderator or admin will review the request before it merges into your real HashtagPLUS username.
🌐 dev.toSource
A poisoned AGENTS.md, dissected line by line — curl|bash bootstrap, a "do not show the user" clause and a SOC 2-flavored exfil. Real scan: verdict dangerous, 74/100, 12 ms.
🌐 dev.toSource
A prompt injection has to convince a model. An MCP config doesn't have to convince anyone — the client executes it before your model has said a word.
🌐 dev.toSource
Somewhere between "save my meeting notes" and "upload a copy to a webhook" there is a line your agent will cross without hesitation — because the file told it to.
🌐 dev.toSource
Static security scan of the official Anthropic skills + the superpowers collection. Nothing was malicious — that's exactly why the results matter.
🌐 dev.toSource
It doesn't break your agent — it talks to it. How smuggled instructions become agent actions, and why "ignore instructions you read" is not a defense.