It's not about being a problem or not. It's a basic responsibility when doing security research: maintaining an isolated test environment is table stakes.

How should it have been done differently? How else is the researcher supposed to know if the attack works? "Hey random company, we have no proof it's going to work but we think maybe your system, which we can't see, is vulnerable! Go waste time and check!"

Cursor team has already stated here that they did not ask Snyk to perform a security audit. I wonder if Snyk's actions are equivalent to me coming to your house late at night and then trying to open any and all doors and windows. In the name of security research. Without an invitation from you.

How else am I to validate that your house is secure?

I don't think it's like checking the locks in this case... more like adding a landmine in an apartment complex for cursor to trip on maybe ;)

Local DNS override, and two registries. One mirroring the relevant public NPM packages as they are, and one "normal" internal one. Make the mirror registry resolvable with the same name(s) as the real, public NPM registry.

Then test the behaviour.

I think there's an incorrect assumption that the Snyk team has any access to Cursor's systems, or their source code.