skill
disk-backup-skills
Agent skills: external disk diagnostics and Rescuezilla backup workflow
Set up
Skills are installed by placing their folder where your agent loads skills.
The download comes from github.com/ssearles/disk-backup-skills itself, not from our server, so what you get is exactly what its author published.
Install instructions come from the README
README
synced Oct 22, 2026Disk & Backup Skills for AI Coding Agents
Two agent skills covering external-disk diagnostics and a Rescuezilla backup workflow, written on Omarchy (Arch/Hyprland) from a real troubleshooting session.
They came out of fixing a genuine problem: a 2 TB backup disk that mounted read-only at every boot after a power cut, losing one backup directory. The disk turned out to be perfectly healthy — the filesystem format was the problem.
Skills
linux-disk-diagnostics
Diagnosing and safely repairing external USB drives.
- Read-only mount triage — three different causes, three different fixes
- SMART interpretation, including why offline self-tests abort on cheap USB bridges and what to use instead
- Full-surface verification with
dd - Safe repartitioning and reformatting
- Filesystem selection for drives shared between Linux and Windows
- Verifying copies without producing false mismatches
rescuezilla-backup-workflow
Running multi-machine disk-image backups.
- Media and filesystem choice
- Drive setup, labelling, identification
- Copying image sets between disks
- Keeping two independent copies in sync with Grsync
- Restoring an image
- Troubleshooting backup-target failures
Why these exist
The session that produced these cost roughly twelve hours, and most of it went to traps that aren't documented anywhere:
| Trap | Consequence |
|---|---|
| Disk letters change between reboots | Risk of wiping the wrong drive |
partprobe makes udisks2 auto-mount | mkfs fails with Device or resource busy |
| A GUI sync tool omits the trailing slash | Creates a spurious wrapper folder |
Parallel xargs reorders checksum output | False "MISMATCH" reports |
xargs -P 4 over one USB bus | 30-minute job takes 6 hours |
| Interrupting a copy-and-delete | Leaves duplicate files |
| Cheap UAS bridges abort SMART self-tests | Test can't be run or observed |
| Image larger than the target disk | Rescuezilla refuses to restore until the partition is shrunk |
| Image smaller than the target disk | Partition left undersized until grown after the restore |
| Omarchy root is LUKS-encrypted | GParted cannot resize it until unlocked with the original login password |
Each is documented with the evidence that identified it.
Restoring onto a differently sized disk
An image taken from one laptop will not restore onto a laptop with a different disk size without a resize first. GParted is on the Rescuezilla flash drive, so no extra download is needed.
| Image vs target | What happens |
|---|---|
| Image larger | Rescuezilla refuses to restore — shrink it first |
| Image smaller | Restores fine, but the partition is left undersized — grow it after |
The passphrase is the Omarchy login password from the machine the image came from. Not a separate disk password, not a LUKS-only password — the same password you use to log in.
This was found and tested by hand, on a real restore, after it looked like GParted had simply hung. It does not hang: Omarchy's root partition is LUKS-encrypted, and GParted will not resize an encrypted container until you unlock it. Select the encrypted partition, click its gear (key) icon, and enter that login password. The icon sits on the partition row rather than in a menu, which is why it is easy to miss.
You need to do this twice when restoring in either direction — once to shrink before the restore, once to grow after it.
How small is safe: both Rescuezilla and GParted warn that not every resize will work, which is true and unhelpful on its own. The rule that actually worked: stay above the size of the coloured band that the installed partition occupies in the GParted disk graphic. Never drag the partition edge below that band. Anything larger than it resized without trouble.
Full procedure in
skills/rescuezilla-backup-workflow/reference.md.
Installation
git clone https://github.com/ssearles/disk-backup-skills.git
cp -r disk-backup-skills/skills/* ~/.agents/skills/
Skills live in ~/.agents/skills/. Each is a folder containing a SKILL.md
with YAML frontmatter — the description field determines when an agent loads
it, so it is written as trigger vocabulary.
If you use Omarchy, ~/.config/hypr/ and ~/.config/omarchy/ can be
git-version-controlled too, so config changes survive a reinstall.
Confidence markers
Claims are tagged throughout:
[PROVEN]— directly observed and diagnosed on real hardware[INFERRED]— concluded from output, not isolated in a controlled test
The distinction matters: some guidance is established by measurement, some is reasoned from symptoms. Treat them accordingly.
Requirements
Linux with smartmontools, util-linux (badblocks), exfatprogs,
e2fsprogs, parted. grsync for the GUI sync workflow. rsync throughout.
Verified on Omarchy with kernel 6.x and smartmontools 7.5, with 2.5" 5400 rpm drives behind JMicron USB-SATA bridges.
Related
Restoring an image leaves the machine carrying the source machine's
machine-id, hostname, and journal. That is a different problem from imaging
one, so it lives in its own repo:
cloned-machine-identity.
Licence
MIT — use, modify, redistribute freely.