Every LLM site looks like Vercel which is such a shame.
Untitled topic
11–20 of 21 posts
Re: undefined
#12For dev-first environments, I just wrote a simple wrapper for 1password https://github.com/na4ma4/1password-direnv-tool >, although I found their application SDK to be quite broken (will stop working after a few hours for no idea why), I didn't write it for anyone other than me, so it works fine for my use case :). I find it super easy to just make a document, and enter key-val details as attributes.
Why not use the op cli? It seems to do the same thing?
Re: undefined
#13I assume other processes running as the same user can still freely read the environment, for example using `ps -Eww` on Mac or inspecting /proc on Linux, right? If so, that's an easy way for a rogue process to bypass the local encrypted vault entirely.
Re: undefined
#14It literally takes 20 minutes to hack this together with an agent and tell it to 'create a stunning landing page'..
Re: undefined
#15This reminds me of the time a trainee dev was doing some work on a security module to hide the IP of the server. Just spending energy on a non threat.
Re: undefined
#16I couldn't find the technique used above the fold (or a short way below). Is this something more (and something more interesting) than just standard spawned process inheriting the parent process environment? IOW is this actually injecting in the true sense of the word? Because that'd be interesting.
Reads like it’s not copying the parent, it’s manually constructing the env dictionary to be passed to execve explicitly. I do this in one of my tools at work because developers were exfiltrating secrets and hand jamming them into .env files.
Something tells me that if I look, I'll see a .env being used in the actual tool
Re: undefined
#17I assume other processes running as the same user can still freely read the environment, for example using `ps -Eww` on Mac or inspecting /proc on Linux, right? If so, that's an easy way for a rogue process to bypass the local encrypted vault entirely.
Yes. Every clone of this idea does the same thing and a new one pops up every week. When I try to point out that the secrets should be exposed through file namespaces instead of ENV vars, the amount of hostility is shocking.
Re: undefined
#18Earlier quoted context omitted.
Reads like it’s not copying the parent, it’s manually constructing the env dictionary to be passed to execve explicitly. I do this in one of my tools at work because developers were exfiltrating secrets and hand jamming them into .env files.
Yeah, so, it's not injecting? To inject something into X, X needs to exist. X does not exist yet when execve is set up. I'm not being pedantic. I just want to read about injection when I'm promised injection :-) because that'd be technically interesting for me. Plainly calling execve isn't so much, I have the manpage here already :-)
[flagged]
Re: undefined
#19Thank you! Hurray! A yet another brand new credentials "local-first" cloud with very transparent and intuitive brand name, that is going to be "responsible" for "decades" (hopefully years) of storing and maintaining someone's credentials with undisclosed infrastructure, legal terms in cases of the credentials leaked during a bug (e.g. at E2EE/secure channel session), and no actual comparison stated between the curren…
[flagged]
Re: undefined
#20This reminds me of the time a trainee dev was doing some work on a security module to hide the IP of the server. Just spending energy on a non threat.
[flagged]