# I asked Claude (Fable 5.2) to audit my whole homelab. It found malware disguised as movies on my NAS. — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/a28343d894eb78a5

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-11-10 23:38 ingested 2026-11-11 02:37

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1xt2co0/i_asked_claude_fable_52_to_audit_my_whole_homelab/)

## retained text



```
I’m a chemical plant operator, not an IT professional. Continuous improvement is part of my job, so I apply it to my homelab too. And since I don’t always know whether a change is safe, the AI agents that act on my machines run under strict rules: a restore point before every change, a guard that blocks destructive commands, nothing irreversible without my explicit approval, and every change goes through a pull request I review.

When Opus 5.5 came out, I asked it for a full audit: servers, websites, scripts, backups, and all the automation that keeps things updated and self-repairing. It found things to improve, but nothing fishy.

With Fable 5.2 on trial, I ran the same audit and asked it to challenge everything built so far. While cleaning an old 2 TB download folder, it first built a read-only inventory and matched every video against my library by exact size. That’s how it caught three fakes that had been sitting there all along:

* two “movies” that were actually Windows executables (`.exe`)
* one “TV episode” that was a 1.1 GB Windows shortcut (`.lnk`)

It never opened them. It flagged them, explained that they’re harmless on a Linux NAS but dangerous with one double-click from a Windows PC on the network share, and waited for my OK.



Same homelab, same kind of audit, very different depth. Nice catch, Fable. 👏
```



document fetched 2026-11-11 00:20 · canonical https://reddit.com/r/ClaudeAI/comments/1xt2co0/i_asked_claude_fable_52_to_audit_my_whole_homelab
