# Has anyone tried using an AI agent to pentest their own homelab's security? — retained record

> Retained reddit record from reddit:homelab.

Canonical: https://highsignal.app/data/records/bf98d2742b1c5650

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:homelab)  published 2026-11-02 12:25 ingested 2026-11-03 02:41

[original source ↗](https://www.reddit.com/r/homelab/comments/1xlvt0f/has_anyone_tried_using_an_ai_agent_to_pentest/)

## retained text



```
With the recent influx of CVEs, and AI agents like Claude Code getting increasingly (and dangerously) better, it looks more and more likely that the typical hobbyist homelab (which probably doesn't have top tier security) might get randomly compromised.

I don't want to just sit still wondering if and when I might get hacked and lose my data until I can restore from limited backups (I don't have an off-site backup of my CCTV recordings for example due to size). I also have HomeAssistant installed, which can control certain things in my house and might pose a fire risk (smart plugs) or at least a nuisance (playing media at 3am) if it gets compromised by a bad actor with ill intentions. I already have everything that gets exposed sit behind a CloudFlare tunnel, but even that got an emergency security patch last week due to the eMudhra certificates thing...

So I got the idea of configuring a Claude agent to intentionally try and compromise my network and devices to see what vulnerabilities it can find and how to patch them. But given how incredibly wrong this can go if Claude isn't given very specific instructions, or if it just decides to go rogue, I'm very skeptical of attempting something like this. Has anyone tried this before?

I'm mostly asking just to get a confirmation that this is a dumb idea, so feel free to bash it. Or maybe I'll get surprised and see some success stories, who knows?

---

**Edit:** would be cool if people stop just mindlessly downvoting the post because you don't agree with it, and actually contribute to the discussion. The whole point of this is to have a discussion about people's past experience with something similar and either how it should be done, or why it should be avoided.
```



document fetched 2026-11-03 00:37 · canonical https://reddit.com/r/homelab/comments/1xlvt0f/has_anyone_tried_using_an_ai_agent_to_pentest
