THIS CLAUDE CODE PROMPT SAVED ME… WHAT WOULD HAVE HAPPENED IF I HADN'T CAUGHT THE HACK ON MY HOME SECURITY SYSTEM?
short version: someone had their own login to my camera system since july. 214 logins in 90 days. 31 were mine
last night I played out the version where I never ran the prompt
> they already had my schedule. 90 days of footage is a calendar of exactly when the house is empty
> they didn't need to crack the alarm. cam 01 has the keypad in frame. they watched the code get typed
> the average burglary costs ~$2,661 (FBI). that's the cheap ending. someone walks in with your code, the alarm logs it as you, no broken window. now explain that to your insurer
> cam 04 sees my desk and my monitor. bank tab, brokerage app, a 2fa code on my phone when I'm careless. a burglar takes the TV. someone with that feed takes the account
> cam 06 is the bedroom. I don't need to explain what that footage is worth to the wrong person
and none of this would have looked wrong. no alert. no broken lock. the app stayed green the whole time
a hack you notice is an incident. a hack you don't notice is a subscription
what I did after (claude changed nothing itself, it handed me the list):
screenshot the logs first. killed UPnP and port 8000. deleted the extra account. new admin password + 2FA. firmware update. and changed the alarm code, because a camera had already seen the old one
if someone had your camera feeds for 90 days, what's the first thing they'd learn about you?
prompt ↓
You're running on my laptop on my home Wi-Fi. Goal: audit MY home security system (IP cameras, NVR/DVR, alarm hub, router) and tell me if anyone besides me has access. Rules: read-only. My network only. Never log into anything, never try passwords, never change a setting, never open a video stream. Ask before installing anything. If a screenshot shows a password, tell me to redact it.
1. Find every device on my network (arp table, ping sweep, mDNS, SSDP, ONVIF discovery). For each: IP, MAC, maker from the MAC prefix. Label the cameras, the NVR and the alarm hub.
2. For each security device: check camera-related ports only (80, 443, 554, 8000, 8080, 37777, 34567), firmware version if it shows, and whether that model still gets updates.
3. Ask me for screenshots of my router's port forwarding and UPnP tables. Flag every rule that points at a camera or the NVR, and anything UPnP opened by itself.
4. Ask me for screenshots from the NVR / camera app: user list, login or access log, active live-view sessions, linked phones. Flag accounts I didn't create, logins from IPs outside my network, logins at times I wasn't using it, and sessions that aren't my devices.
5. Ask me what each camera can see: door keypads, desks and screens, a safe, when the house is empty. If anyone else had access, list every code, PIN and password a camera could have seen, so I can change them.
6. Output: one self-contained HTML file with a 3D map (three.js) of my home, every device pinned, anything suspicious in red. Plus fix.md: what you found, how sure you are (high / medium / low), and fix steps in order. Step 1 is always: screenshot the evidence before changing anything. Change nothing yourself.
THIS CLAUDE CODE PROMPT CHECKS YOUR WHOLE MAC FOR MALWARE IN ~5 MINUTES. THE SCARIEST THING IT FINDS IS NEVER A VIRUS
antivirus looks for files it already knows are bad
this flips it. one question: what on this machine can start by itself, talk to the internet or read your screen, and can you explain every single one
what the agent goes through:
> everything that launches on its own: login items, launch agents, background items
> every browser extension in every profile, with its real permissions
> configuration profiles. on a personal mac the right number is zero
> apps that aren't signed or notarized + anything installed in the last 30 days
> every live connection, mapped to the app that owns it
read-only the whole way. it deletes nothing and asks before any sudo
and the find that matters is almost never a virus. it's a coupon extension from 2 years ago with permission to "read and change all your data on all websites"
your bank's website included
malware has to sneak in. that one you installed yourself
the prompt (claude code, your own mac):
Audit this Mac for anything I can't explain. Read-only: never delete, uninstall or change a setting. Ask before installing any tool. If a command needs sudo, print it and let me run it.
1. Basics: macOS version and pending updates, FileVault, firewall, Gatekeeper.
2. Everything that starts by itself: Login Items, LaunchAgents and LaunchDaemons (user + system), background items (print the sudo sfltool dumpbtm command for me).
3. Configuration profiles and device management. I should have none unless my employer manages this Mac.
4. Browser extensions in every browser and profile, with their permissions.
5. Apps that aren't signed or notarized, and anything installed in the last 30 days.
6. Live outbound connections (lsof -i -nP), each mapped to the app that owns it.
7. Report only what I can't explain: unsigned, unknown, recently added or phoning home. High / medium / low confidence on each. Remove nothing.