I've always had a tendency to turn good ways of working into something reusable.
Before AI, that meant bug templates, PRD structures, meeting-note formats, tickets — anything that stopped me or the team from figuring out the same process from scratch every time.
Over the last few months, I've been doing the same thing with Claude Skills.
Most started with a real task. I'd use Claude to help me do something once, refine the process until it worked well, then turn that workflow into a Skill the team could reuse.
Writing PRDs
Instead of asking Claude to “write a PRD,” I built a workflow around how I actually approach one: understand the project, read the source material, separate what's decided from what's still unclear, ask questions, then draft. The Skill captures that process rather than just generating a document.
Turning designs into Jira tickets
This started with using Claude to help break designs into development work. I refined it so it gathers the surrounding project context, reads the designs, proposes how the work should be split, and waits for approval before creating anything.
Keeping project information in sync
I also turned some of the less interesting maintenance work into Skills — things like keeping development trackers aligned with Jira and pulling scattered project information together for case studies or blog posts.
None of these are particularly flashy uses of AI. That's partly why I find them useful.
The goal isn't to hand over the product decisions. It's to make the repetitive parts around those decisions easier: gathering context, structuring information, spotting what's missing, and turning decisions we've already made into something useful.
It's the same approach that had me making templates and databases pre-AI:
When something works, make it reusable.