The One Skill That Flipped the Script on AI Coding (And Why You Need It)

You’ve probably noticed it: the endless loop of asking AI to write code, watching it guess wrong, then rewriting the prompt, again and again. It feels like a toxic relationship. You’re doing all the work, and the AI is just… guessing.

Now imagine a skill that has been installed over 600,000 times — ranking #3 in the global skill installation charts — and its core content is only a few lines of text. A few sentences. That’s it.

This is grill-me. And it doesn’t make the AI smarter. It makes you smarter.

Let that sink in for a second.

We’ve been told the future of programming is about writing better prompts. Better instructions. More precise context. But the biggest breakthrough in AI coding is a skill that forces you to stop writing prompts and start answering questions. The AI doesn’t guess anymore. It interrogates you.

Here’s how it works. You install grill-me with a single command: npx skills add https://github.com/mattpocock/skills --skill grill-me. Then, in any AI coding tool like Cursor, you type a vague request — something like ‘build me a tool to manage my AI skills’ — and hit /grill-me.

What happens next is magic. The AI doesn’t start coding. It starts asking questions. One at a time. It drills down into your requirements, your assumptions, your technical choices. It follows a decision tree: each question opens a new branch, and it won’t move to the next until you answer. It asks you for your recommendation on every decision. It checks your project files for facts but never makes a judgment call without your input.

You end up in a conversation that feels like pairing with a world-class architect, not a code monkey. You say ‘I want a web app’ and it asks ‘What framework? React or Vue? What about state management? Do you need a database?’ You say ‘I don’t know’ and it says ‘Here’s my recommendation: use SQLite for simplicity.’

After fifteen rounds of this, you’ve done something remarkable: you’ve designed a complete system before a single line of code was written. You’ve turned a vague idea into a blueprint, and you barely lifted a finger.

This is the insight that makes grill-me viral. The real bottleneck in AI programming isn’t the AI’s coding ability. It’s the human’s ability to articulate requirements. We think we know what we want, but we don’t. We hand the AI a half-baked idea and expect it to read our minds. Then we get frustrated when it builds something useless.

Grill-me flips the script. It forces the human to do the hard work of thinking before the AI does the hard work of coding. The result? Less rework, fewer surprises, and a finished product that actually matches your intent. In one 10-minute session, I built a fully functional desktop app that would have taken me hours of back-and-forth with a traditional AI workflow.

This isn’t just a productivity hack. It’s a fundamental shift in how we think about human-AI collaboration. The best prompt engineering isn’t about making the AI smarter. It’s about making the human clearer. Grill-me is the rubber duck debugging method on steroids — a duck that asks questions, offers recommendations, and checks your codebase for facts.

If you’re still writing prompts and hoping the AI gets it right, you’re doing it wrong. The future of AI coding isn’t about telling the AI what to do. It’s about letting the AI tell you what you actually want.

Try it. The 600,000 people who already have can’t be wrong.