The Backup You Don’t Have Is the One That Kills Your Q4
Every cross-border operator I know runs the same quiet gamble: brand assets, supplier contracts, compliance docs, and ad creative live in Google Drive, and the local copy is whatever the sync client last felt like doing. We obsess over inventory reconciliation and ad attribution, then trust a background daemon with the only copy of our trademark filings. So when a solo developer launches a Linux desktop client for Google Drive — Skymir — because he lost 1,437 Google Docs to an rclone bisync gone wrong, I don’t read it as a Linux story. I read it as a file-integrity story, and file integrity is the least-glamorous, highest-leverage problem in this business.
What Skymir Actually Solves (And Why the Linux Framing Is a Distraction)
The maker is David Avellan, a solo developer who switched to Linux and hit the wall anyone who has tried to run an agency or seller workstation on Linux eventually hits: Google ships a desktop Drive client for Windows and Mac, but never for Linux, so the only way to touch a Google Doc is through a browser tab. His prior workaround was an rclone bisync setup, which he describes as having deleted 1,437 of his Google Docs because it treated cloud-only Workspace files as missing from disk. That’s the failure mode worth understanding: rclone is a general-purpose sync engine, and a general-purpose sync engine pointed at a folder where “not present locally” and “deleted” look identical will eventually make the wrong call at scale.
Skymir is a native GTK desktop app that mirrors selected Drive folders to local disk, handles Google Docs as clickable shortcuts rather than trying to materialize them as files, and lives in the system tray. It’s a 19 MB download. Saves reach Drive within seconds; changes from Drive arrive within about a minute; and deletes route to trash rather than permanent removal. There’s a 14-day free trial with full features, no account or credit card required, and per one commenter’s observation, the pricing model is a one-time payment rather than another monthly subscription — a detail the maker confirmed with a wry “one less thing to pay monthly.”
The install path and landing page are at skymir.app/install and skymir.app respectively, with support at [email protected].
The safety rule is the actual product
Buried in the comment thread is the feature that matters more than the GTK wrapper: a rule that refuses any sync operation that would delete more than half your files. Commenter Uddipta Mahanta called it out directly, and the maker confirmed it’s adjustable from 0–100% in settings. Mathew James flagged the same thing as “a thoughtful touch.” I’d go further: this is the only feature on the page that maps to a real operational risk, and it’s the reason I’m writing about a Linux utility on a cross-border e-commerce blog at all.
How It Compares to What You’re Probably Running Today
Let me be blunt about the incumbent landscape, because most sellers reading this are running one of four things and none of them are great.
The web interface. Free, zero install, and the default for anyone who doesn’t want to think about sync. It’s also the reason your creative team is re-downloading the same product photo pack every time a new VA joins, and why nobody can grep across 400 supplier PDFs at once. Browser-only means no local search, no bulk operations, no scripting.
rclone. The power-user answer, and genuinely excellent — until it isn’t. The maker’s own 1,437-doc incident is the cautionary tale: rclone will do exactly what you configure it to do, including treating cloud-only files as deletions. If you’re running rclone bisync against a Workspace folder and you haven’t tested your restore path in the last 90 days, you’re one misconfigured flag away from a very bad Monday.
Insync. The long-standing commercial option for Linux Drive access, and the one the maker explicitly names as a comparison point when asking what features users would need before switching. It’s mature, but it’s a subscription, and for a lot of operators the calculus on yet another monthly line item has gotten ugly.
Native clients on Windows/Mac. Fine, if your whole team is on those platforms. The moment you hire a developer, a data analyst, or anyone who prefers Linux, you’ve got a two-tier file access situation and a support burden.
Why Amazon sellers should care more than Shopify ones
Here’s my read on who actually needs this. Pure Shopify DTC operators live inside a browser all day — Shopify admin, Klaviyo, Meta Ads Manager. Their file needs are mostly creative assets, and those often live in Dropbox or Frame.io anyway. The pain is real but diffuse.
Amazon FBA brand owners are a different animal. You’re dealing with supplier contracts, compliance certificates, test reports, brand registry documentation, A+ content source files, and — critically — reimbursement claim evidence. Amazon’s Seller Central case logs are a graveyard of “please re-upload the invoice” threads. If your documentation folder is browser-only, you’re doing every dispute the hard way. A mirrored local folder that you can search, batch-rename, and back up to a second location is worth more to an FBA seller than to almost anyone else in e-commerce.
Same logic applies to TikTok Shop and Temu sellers juggling platform-specific compliance docs, and to Etsy shops where the product is the digital file.
Where the math breaks
I want to be honest about the limits of the safety rule, because “refuses to delete more than half your files” sounds bulletproof and isn’t.
If you have 10 files in a folder and legitimately delete 6, the guard trips and blocks a correct operation. If you have 10,000 files and a bad sync wants to delete 4,900, the guard happily lets it through. The threshold is percentage-based, not absolute, and percentage-based guards are always wrong at one end of the distribution. The maker notes it’s adjustable from 0–100%, which is the right escape hatch, but it also means the protection is only as good as the operator who configured it. Set it to 100% and you’ve disabled it. Set it to 10% and you’ll be fighting false positives on every cleanup.
The correct posture: treat this as a seatbelt, not a roll cage. It reduces the blast radius of the most common failure mode. It does not replace version history, and it does not replace an independent backup of anything you’d cry about losing.
Why the one-time pricing is a bigger deal than it sounds
Magdalena Anderson flagged the one-time payment as “a nice change,” and I think she undersold it. Cross-border operators are drowning in SaaS subscriptions — Helium 10, Jungle Scout, a repricer, a review tool, a returns platform, three ad tools, a 3PL portal, a freight forwarder dashboard. Every one is $50–$500/month. A utility that does one job and doesn’t add to that stack is genuinely differentiated, and it’s a pricing philosophy more indie tooling should copy.
The tradeoff is obvious and worth naming: one-time pricing on a solo-dev product means the sustainability question is real. Who funds the Google API changes, the GTK updates, the Linux distro compatibility matrix in year three? Not disclosed, and it’s the thing I’d want answered before standardizing a team on it.
What Cross-Border Sellers Should Actually Steal From This
Strip away the Linux specifics and there are three transferable lessons here that have nothing to do with which desktop OS you run.
One: audit your sync topology before it audits you. The maker’s rclone failure wasn’t exotic — it’s the standard failure mode of any bidirectional sync pointed at a folder with cloud-only placeholders. If you run Dropbox, OneDrive, Google Drive, or a NAS sync across your ops team, you have the same exposure. Test your restore path this quarter. Actually delete a file, actually restore it, actually verify the version you got back is the one you wanted.
Two: destructive operations should require a threshold, not a confirmation dialog. Everyone clicks through confirmations. Nobody clicks through a hard block. If you’re building internal tooling — bulk listing updates, inventory adjustments, price changes across a catalog — steal the “refuse if this exceeds X% of the set” pattern. It’s a five-line guard that prevents the class of error you can’t undo. A bulk price update that touches 80% of your SKUs should probably require a second human, not a second click.
Three: the boring utility layer is where operator leverage hides. Nobody writes LinkedIn posts about file sync. Everybody writes LinkedIn posts about the AI agent that 10x’d their ad creative. But the operator who never loses a compliance document beats the operator with the flashiest stack, every single time. When you’re evaluating tooling spend, the unsexy infrastructure line items deserve more scrutiny than they get.
A note on the “solo dev builds the thing Google wouldn’t” pattern
Julian Ting’s comment — that Google not making a Drive client for Linux still surprises me — captures something structural. Platform vendors abandon niches, and niches get served by individuals with domain pain. That’s the same dynamic that produced half the tools in your stack: someone got fed up with Amazon’s reimbursement process and built a claims tool, someone got fed up with Shopify’s native returns flow and built a returns portal.
The upside is speed and focus. The downside is bus factor. When you adopt a solo-dev tool into an operational workflow — not a nice-to-have, but a “this is how our docs sync” workflow — you’re taking on concentration risk. That’s not a reason to avoid it. It’s a reason to keep the exit ramp clean: know where your files actually live, know how to get them out, and don’t let any single utility become the only path to your own data.
What I’d Watch / Test Next
This week, three concrete moves.
First, if you’re on Linux or have Linux users on your team, grab the 14-day free trial — no account or credit card required — and point it at a non-critical Drive folder first. Watch what happens when you delete a file locally. Confirm it lands in trash. Then watch what happens when you delete from the web. Confirm it propagates. Don’t start with your compliance folder.
Second, regardless of OS, run a restore drill on whatever sync tool you currently use. Pick a folder with 50+ files, delete three, restore them, verify contents. If you can’t complete that drill in under ten minutes, you have a problem you don’t know about yet.
Third, go read the maker’s landing page with the specific question he asked in mind — does it make the app clear quickly? — and then ask the harder version for your own stack: if a new VA joined tomorrow, could they figure out where your brand assets live, how they sync, and what happens when something gets deleted? If the answer is “they’d ask me,” you’ve found your next infrastructure project.
The feature I’d actually push for, if the maker is reading: an absolute-count threshold in addition to the percentage one. “Block if this deletes more than 500 files, period” catches the large-catalog case that a percentage guard waves through. That’s the version I’d standardize a team on.
