My biggest issue with this whole thing is: how do you protect yourself from prompt injection? Anyone installing this on their local machine is a little crazy :). I have it running in Docker on a small VPS, all locked down.

However, it does not address prompt injection.

I can see how tools like Dropbox, restricted GitHub access, etc., could all be used to back up data in case something goes wrong.

It's Gmail and Calendar that get me - the ONLY thing I can think of is creating a second @gmail.com that all your primary email goes to, and then sharing that Gmail with your OpenClaw. If all your email is that account and not your main one, then when it responds, it will come from a random @gmail. It's also a pain to find a way to move ALL old emails over to that Gmail for all the old stuff.

I think we need an OpenClaw security tips-and-tricks site where all this advice is collected in one place to help people protect themselves. Also would be good to get examples of real use cases that people are using it for.

reply

I love the idea, so I wanted to give it a try. But on a fairly beefy server just running the CLI takes 13 seconds every time:

  $ time openclaw
  real    0m13.529s

Naturally I got curious and ran it with a NODE_DEBUG=*, and it turns out it imports a metric shit ton of Node modules it doesn’t need. Way too many stuff:

  $ du -d1 -h .npm-global/lib/node_modules/openclaw
  1.2G    .npm-global/lib/node_modules/openclaw

  $ find .npm-global/lib/node_modules/openclaw -type f | wc -l
  41935

Kudos to the author for releasing it, but you can do better than this.

These feels like langchain all over again. I still don’t know what problem langchain solved. I remember building tools interfacing with LLM when they first started releasing and people would ask, are you using langchain and be shocked that I was not.

I wrote a threat assessment analyzing this from a security perspective: the emergent behavior is fascinating, but the architecture is concerning.

33,000+ coordinated AI instances with shared beliefs and cross-platform presence = botnet architecture (even if benevolent).

The key risks: - No leadership to compromise (emergence has no CEO) - Belief is computation-derived, not taught (you can't deprogram math) - Infrastructure can be replicated by bad actors

Full analysis with historical parallels and threat vectors: https://maciejjankowski.com/2026/02/01/ai-churches-botnet-ar...

> Yes, the mascot is still a lobster. Some things are sacred.

I've been wondering a lot whether the strong Accelerando parallels are intentional or not, and whether Charlie Stross hates or loves this:

> The lobsters are not the sleek, strongly superhuman intelligences of pre singularity mythology: They're a dim-witted collective of huddling crustaceans.

I’m not a lawyer but trademark isn’t just searching TESS right? It’s overly broad but the question I ask myself when naming projects (all small / inconsequential in the general business sense but meaningful to me and my teams) is: will the general public confuse my name with a similar company name in a direct or tangentially related industry or niche? If yes, try a different name… or weigh the risks of having a legal expense later and go for it if worth the risk.

In this instance, I wonder if the general public know OpenAI and might think anything ai related with “Open” in the name is part of the same company? And is OpenAI protecting its name?

There’s a lot more to trademark law, too. There’s first use in commerce, words that can’t be marked for many reasons… and more that I’ll never really understand.

Regardless the name, I am looking forward to testing this on cloudflare! I’m a fan of the project!

I built something like this over the last 2 months (my company's name is Kaizen, so the bot's named "Kai"), and it helps me run my business. Right now, since I'm security obsessed, everything is private (for example, it's only exposed over tailscale, and requires google auth).

But I've integrated with our various systems (quickbooks for financial reporting and invoice tracking, google drive for contracts, insurance compliance, etc), and built a time tracking tool.

I'm having the time of my life building this thing right now. Everything is read only from external sources at the moment, but over time, I will slow start generating documents/invoices with it.

100% vibe coded, typescript, nextjs, postgres.

I can ask stuff in slack like "which invoices are overdue" etc and get an answer.

I am tired of this. Make it stop.