Hacker News new | ask | show | jobs
by nemomarx 4 hours ago
What's poisoned? There's a disclaimer that coding agents shouldn't touch it and some prompt injection stuff that honestly AIs should have defenses for already before you're letting them work with third party code. Nothing really gets damaged?

Even if the prompt actually did work it would just stop the agent from implementing this specific testing framework, which is on the level of making your library incompatible with another or something.