I just shipped something I've been thinking about for a while. Working at an agency, I see the same problem over and over: clients come to us wanting "an AI solution" but they have no idea what prompt engineering actually involves. They think it's like writing an email. Then we spend two weeks iterating on something that should take three days, and the budget blows up.
So I built a starter kit. It's a collection of structured prompts, templates, and a workflow for taking a client brief and turning it into something actually usable. Includes examples for common use cases we see (customer support automation, content generation, data extraction), plus documentation on what to test, how to measure quality, and when to push back on scope creep.
The whole thing is on GitHub right now, free. I'm not trying to monetize it, honestly. I just want to raise the baseline for how people think about this work. When clients understand that prompt engineering is actual engineering, that it requires testing and iteration and sometimes you need to restructure your data, the conversations go so much better.
It's not revolutionary. It's just... practical. Things I wish I had five years ago when I first started doing this stuff. Some of the templates pull from Anthropic's guides and other resources I've found useful, and I've added a bunch of our own learnings from actual client work.
If you're doing agency work and dealing with prompt stuff, I'd love to know if it helps. Or if you think something's missing. I plan to update it as I learn more (which, at the pace this field moves, is basically weekly).
Here's the link if you want to check it out.