Fair concern — that's the most common failure mode for extensions like this.
A few things I've built in to reduce breakage:
First, the parser doesn't rely on a single fragile selector chain. It targets multiple fallback identifiers per element type, so if Google renames one class, the others still catch it. It's not bulletproof but it degrades gracefully rather than exporting blank files.
Second, I'm running a lightweight automated check that pings the extension against a fixed Gemini conversation daily. If the output diff crosses a threshold, I get alerted before users notice.
Third — honestly — the extension is small enough that I can ship a patch within 24–48 hours of a breaking change. That's one advantage of not having a 6-week release cycle.
It will break at some point. I'd rather be upfront about that than oversell stability. The goal is to fix it fast when it does.
