Vibe coding is fast, but it often skips the fundamentals: project structure, typing, testing, error handling, and review discipline.
Matt Pocock's skills repo https://github.com/mattpocock/skills is one of the most practical attempts to fix that. It doesn't just say "write tests" — it gives you repeatable, context-rich engineering prompts you can drop into your workflow.
What makes it useful:
- It turns tribal knowledge into shareable skills
- It enforces engineering rigor without slowing down iteration
- It's framework-agnostic enough for real teams
If you're using AI to ship quickly, the bottleneck is no longer generation speed — it's whether the output is actually maintainable. Skills like these are how you close that gap.
Have you tried structuring your AI-assisted workflow around reusable engineering skills?
