I’ve always avoided typing secrets on the CLI, temporary environment variables is my usual goto.
GenAI and security seem to be things that are a bit on collision course.
I don’t understand why people trust LLMs.
I trust the summaries on my searches as much as I already know or can verify, like any tool, but I don’t log into them. I use the anon ones like duck.ai and lumo and if it limits my usage then time to go anyway. If they track me somehow it doesn’t really matter since I do no harm and come in peace.
Listening to execs and end users that I support at work, their reasoning has been “its a big company with insurance around my data, if they lose it I’ve got a law suit” and I’m just smh
True,especially they (usually) train on everything.
They don’t know how any of it works, and have no reason to ever learn now that they have a magic box, so of course they’re going to do whatever the magic box tells them as long as the magic box gives them what they want.
This. The whole selling point is that it does everything and you don’t need to worry about anything.
Claude Code is not a LLM. It’s a regular application communicating with an LLM. It might be very vibe coded, seeing who it is written by, but it’s not an LLM that’s the problem here.
It’s just bad application design.
The bad application design here is allowing an unreliable LLM write access to your repo and to the internet.
And absolutely the LLM is the problem.
You know the LLM gets stuff wrong, yet you put it in charge. Sooner or later, it’s inevitably going to go very, very wrong.
People were wrong to trust it, people read all the stories of it going wrong, and I don’t understand why anybody still does trust LLMs.
It always was a decision to outsource your most valuable thinking to a subscription controlled by companies who were so obviously on the initial stages of an enshittification journey. Why else were they pushing it so very hard?!
“Yeah, sure, some of its towers fall down, but look how fast it builds them, and look at how heavily below cost the initial subscription is! Let’s replace all our institutional knowledge and experience with a random sentence generator that ate stackoverflow! Yay!”
I don’t understand why people vote for narcissistic, fraudulent, criminal, child-raping bullies, either. It was so clearly and obviously going to go wrong.
I genuinely don’t understand how is everyone having such a bad experience. Granted, I hardly chat or use it for research, but every task I’ve given to codex has been successfully fulfilled. Two last ones were optimizing a web page that drew entire world map with lots of details (actually blew my mind, it was like 95% faster) and fixing a mod in a videogame. I’ve created custom app for a radio station I loved but didn’t have basic features like previous songs, I’ve created an entire web game in 3 hours, I’ve created mods for games, etc., all without a fail.
This thing has saved me lots of headache, yes everyone here claims it’s so bad and useless and a complete bubble, meanwhile in the last year it has basically only started working and only now can businesses start adapting it (I’m sure early adaptors were not happy)
EDIT: you just know this comment will receive silent grunts and hardly any real counter-arguments because I’m right.
Absolutely it’s got a lot better, and it can do a lot of things, but it can also oopsie wipe your entire repo (and be very apologetic afterwards). I can’t understand how you’ve not heard those stories and I can’t understand how you would give one executive power if you had.
@davidagain Short answer: LLMs “push all our Darwinian buttons” (Sherry Turkel), the ones that make us social beings.
Doesn’t this all come back to bad code review though? Laike if I saw this in a commit I’d kill the PR and demand they rotate the password now. And also add this to the gitignore
oooh time to pull back the image again :3
Devs accidentally committing secrets and forgetting to gitignore env files is nothing new, this is just a new surface. It’s why secret scanning is such an important part of a CI/CD pipeline.
a high entrop for any data going out from that blind CLI should be banned without exception. The user then needs to approve. its the responsibility of Anthropic.
I can only speak anecdotally, but from what I’ve seen firsthand the user does need to approve. In fact, the user needs to approve almost everything, unless they choose to put the CLI into “auto-mode” which should just be assumed is at one’s own risk.
In regard to
settings.local.jsonI’ve only ever seen it automatically created gitignored. I’m not saying it’s not possible for the user to prompt in some way to intentionally or accidentally bypass that, but still.At some point this honestly isn’t Anthropic’s problem. And a developer that takes the stance of “privacy and security is the responsibility of my toolkit” is one that is either very green, or reckless.
it’s a bit tedious in some ways but I really liked working with GitHubs secret system. As long as you ensure nothing changes in the GitHub run system you could use various key signing mechanisms or even uploads to service providers like AWS with out sharing via dot files or user configuration only files that end up being committed to the repo.
What the fuck is this website?
Are they ripping off HN and AI summarising their posts our what
https://apiiro.com/blog/coding-agents-leaking-secrets-github-dont-ask-again
is this the original?
This is what I thought people referenced for over a decade whenever they mentioned HN because no one pointed me to the actual HN.
y combinator was founded by sam altman and is really mostly AI schill and very boring nowadays
That explains why LLMs are so stochastic.
The slop is annoying but it’s getting better.
I didn’t look at hn for half a year because every post was about AI. But now interesting things are popping up again.
Looks at top posts this morning
Yeah, no fam lol
Yeah but at least the top 10 posts also contain non-ai stuff usually
You’re tracing to justify your response lol
And you initially replied with something that had little to do with my response?
The “nothing new, devs always commit secrets” take misses the bit that makes this one nasty: the file isn’t a .env someone chose to commit, it’s an agent settings file that silently accumulates the exact commands you approved, secrets included, and then gets accidently committed with the project because it lives in the repo, so it ends up publically exposed. YSK that’s a different failure mode to a one-off .env leak.
If you use Claude Code (or any coding agent, tbh), the practical fix is boring but it works: rotate anything that was in that file — deleting the file from the repo doesn’t un-expose it, it’s already scraped — then check git history, because the secret is still in there unless you rewrite it. Add settings.local.json to .gitignore, run a secret scanner like gitleaks in pre-commit, and stop using ‘don’t ask again’ for commands that echo tokens or passwords. Also audit the approved-commands list now and then, and keep tokens scoped as tightly as you can so a leak like this is annoying rather than catastrophic. FYI, GitHub’s secret scanning will catch some of this after the fact, but it’s no substitute for not committing the file in the first place.
Fuck off, clanker.
I don’t like it when bots write comments pretending to be human.
I also don’t like it when humans become meat proxies.
