linux-security-hardeninglisted
Install: claude install-skill SpeakingColors/claude-skills
# Linux security hardening (Bazzite / Fedora Atomic + home basics)
Labels: VERIFIED (source) = read in a fetched page; COMMUNITY = widely repeated, not confirmed in a fetched page; UNVERIFIED = unconfirmed, test before trusting.
## How to teach (assume the reader is not a programmer)
- One change at a time. Plain-words explanation first, then the exact command, then how to check it worked, then how to undo it.
- Never paste several commands at once. Ask for the output before the next step.
- Never ask for passwords, keys or real IPs in chat. Use placeholders.
- Check the current docs before stating a Bazzite-specific fact; Bazzite changes fast.
## Threat model by tier (plain words)
- Tier 1, stolen PC or drive: someone has the hardware. Defence: full disk encryption plus a secret needed at boot.
- Tier 2, malware: a bad download or app runs as you. Defence: trusted sources only, sandboxing, updates, backups.
- Tier 3, network attacker: someone on the internet or Wi-Fi. Defence: nothing listening (no SSH), firewall on, router basics.
- Tier 4, account takeover: someone logs in as you on a website. Defence: password manager, unique passwords, 2FA, phishing awareness. This is the most likely tier.
## What immutability already gives (and does not)
- System files are replaced as a whole image and you can roll back: VERIFIED (docs.bazzite.gg rolling back page: `rpm-ostree rollback`, two deployments kept by default).
- Updates are image builds that apply on reboot: VERIFIED (doc