So a few weeks back I was using Claude to help write an eBook — content came out great, took like 20 minutes. Then I tried to actually turn it into a proper ePub file.
Big mistake. Spent an entire weekend fighting broken table of contents, margin glitches, HTML/CSS validation errors that Amazon KDP kept rejecting. The AI wrote a whole book faster than my formatting pipeline could package one chapter.
At some point I just stopped and thought — why am I doing this manually when the AI could just... do this too?
So I packaged it as a Skill (not another SaaS dashboard, I have enough of those already) that plugs straight into Claude, Gemini, Cursor, whatever you're already using. You give it a topic or rough notes, it handles the chapters, the ePub 3 packaging, the Kindle PDF, the TOC, the metadata — all in one pass, inside the tool you already have open.
Threw it up on Gumroad for $9.99 lifetime (bumping it up weekly since it's early) mostly to see if other builders hit the same wall I did.
It's now generated 10k+ books for people, which honestly surprised me more than it should have.
Still figuring out:
whether $9.99 lifetime undersells it or if that's exactly why people are trying it
if I should build native integrations vs staying "just a skill you load in"
If you've built AI-powered digital products before, curious how you thought about pricing something like this.
🔗 oneclickebook.vercel.app
This is a good example of packaging a repeatable workflow instead of just sharing a prompt. The hard part, in my experience, is making the skill output easy to verify: source checklist, acceptance criteria, and a final artifact that a non-technical user can actually run. If you add one section, I’d make it a validation gate before export.