I run Claude Code with --dangerously-skip-permissions flag, giving it full system access. Let me show you a new way of approaching computers.

Dixit Ram

Browsers

Community Picks is a section on daily.dev where our community members share the most interesting and valuable content they've discovered online. From insightful articles to handy tools, every post is a gem curated by our dedicated coomunity. To contribute to Community Picks, you need to have at least 250 reputation points, ensuring that only active and trusted members can share their finds.

Community Picks

A developer shares their experience using Claude Code with full system permissions for two months, demonstrating how it transformed their workflow from simple coding assistance to complete system automation. The setup handles everything from content migration and code commits to system configuration and machine setup, saving approximately an hour daily despite potential security risks.

Claude Code is My Computer

From ‘AI assistant’ to everything terminal #

“potential” security risks, riiight

Just wanted to chime in here as a Warp user, you can grant blanket trust in Warp. It’s in the AI settings, under permissions. You have four permissions for fine-grained controls. One of which is “Execute commands” that you can set to always allow.
I personally would never do that. I use Warp as my AI agent of choice. But given the risks of LLMs hallucinating, I would not want it to execute something that I cannot read and understand myself. That might seem counterintuitive to the idea of an AI agent, but I would rather be cautious that sorry.

Honestly I think this is probably going to be more and more common.
There is a massive security problem, but so is there in uploading all your photos to facebook. And we as a global community fell for that hook line and sinker.
So when you tell normies, you can automatically find every picture of a dog on your computer and move it all to a single folder by just saying that… and it’s going to orchestrate a call to a public dog-classification API, and the move files based on the output across your whole machine.
They are not going to care about ‘muh security’ as long as the first 10k or so prompts don’t “rm -rf /” theyl’l be fine, and if it does that every few years accidentally, well that’s no different from a windows machine bluescreening, and refusing to boot which was normal in say early 2000s.

This is exactly the kind of trade-off we keep making, and frankly, it’s a little frustrating. Yes, giving an AI agent the power to execute commands directly is a massive security risk. But let’s be real, how different is that from the blind trust we handed over when we uploaded every single photo, contact, and personal detail to Facebook? We fell for that hook, line, and sinker, and here we are.
Are we really learning? Or are we just repeating the same mistakes with shiny new tech? I’d love to hear from others: Do you think users will finally start demanding better safeguards, or is convenience just too seductive to resist? Because from where I stand, we’re barreling toward more automation with less scrutiny, and it’s going to get messy if we don’t start having these hard conversations now!