Grill Me Skill: The Deceptively Simple AI Prompt That Revolutionizes Code Design
What is the Grill Me Skill?
The Grill Me Skill is an innovative AI productivity tool created by Matt Pocock, a renowned software engineer and educator, as part of his comprehensive "Skills for Real Engineers" repository on GitHub. This powerful skill transforms AI assistants like Claude into relentless interviewers that systematically question and stress-test your software plans, designs, and architectural decisions until every branch of the decision tree is fully resolved.
The Deceptively Simple Prompt: Why These Few Lines Pack a Powerful Punch
At first glance, the Grill Me skill appears almost trivially simple. Here's the complete prompt:
Interview me relentlessly about every aspect of this plan until
we reach a shared understanding. Walk down each branch of the design
tree resolving dependencies between decisions one by one.
If a question can be answered by exploring the codebase, explore
the codebase instead.
For each question, provide your recommended answer.
Just 6 lines. But why is this so powerful?
"Interview me relentlessly" — The Power of Persistence
The word "relentlessly" is crucial. It instructs the AI to not stop after one or two surface-level questions, but to continue digging deeper and deeper. This single word transforms a casual conversation into a comprehensive design review that can last 45 minutes or more, uncovering requirements you didn't even know existed.
"Until we reach a shared understanding" — The Goal-Oriented Approach
This phrase sets a clear termination condition. The AI doesn't just ask random questions—it has a specific goal: achieving mutual comprehension between you and the AI about every aspect of the plan. This prevents the conversation from becoming aimless or superficial.
"Walk down each branch of the design tree" — Systematic Coverage
This instruction ensures the AI doesn't skip important considerations. By explicitly mentioning "each branch," the prompt forces comprehensive exploration of all possible paths, edge cases, and decision points in your design. It's the difference between casual brainstorming and rigorous architectural review.
"Resolving dependencies between decisions one by one" — Sequential Logic
This phrase is brilliant because it prevents the AI from overwhelming you with 20 questions at once. Instead, it tackles dependencies sequentially, building understanding incrementally. This makes complex planning sessions manageable and ensures each decision is made with full context.
"Provide your recommended answer" — Accelerated Decision Making
This recently added line is a game-changer. Instead of you having to explain obvious choices, the AI suggests what it thinks is the best answer. You can simply say "yes" to good recommendations, dramatically speeding up the conversation while maintaining depth.
Real-World Impact: Before and After
One developer on Reddit shared their transformation after adopting the Grill Me skill:
| Before Grill Me | After Grill Me |
|---|---|
| Switch to plan mode | Run /grill-me with feature request |
| Give a prompt | AI asks questions one by one |
| AI might ask 1-2 questions (sometimes none) | Make decisions together progressively |
| AI generates a plan with wrong assumptions | Reach shared understanding before coding |
| Endless revision cycles | 90% accuracy on first implementation |
The developer noted: "By enforcing iterative clarification before plan generation, the grill-me skill appears to reduce downstream rework and improve alignment."
Common Mistakes to Avoid
Matt Pocock identified critical mistakes developers make with Grill Me:
Trying to Answer High-Fidelity Questions
Not all questions are "grillable." Low-fidelity questions like "What URL should this route live on?" work perfectly in grilling sessions. But high-fidelity questions like "How should this UI feel?" require prototypes.
Solution: Use the handoff pattern: grill → prototype → grill again.
Choosing Too Large a Scope
When scope is too large, you hit two problems:
- Hidden high-fidelity questions emerge
- You exceed the model's context window (~120k tokens), entering the "dumb zone" where AI performance degrades
Solution: Break large scopes into smaller, grillable chunks before starting.
Why This Matters for Modern Development
The Grill Me skill represents a paradigm shift from reactive debugging to proactive design validation. Traditional development often means writing code first, then discovering problems. Grill Me flips this: you discover problems during a 45-minute conversation, saving hours or days of refactoring later.
Matt Pocock describes it as the digital evolution of rubber ducking—the practice of talking through your code with an inanimate object to discover flaws in your logic. But unlike a rubber duck, the AI asks intelligent follow-up questions and provides expert recommendations.
Beyond Coding: Universal Application
Interestingly, Pocock uses Grill Me for non-coding decisions too, like figuring out what course to build next. The skill works for any complex decision-making process where you need to explore all angles before committing.
How to Get Started
Install the skill using:
npx skills@latest add mattpocock/skills
Then run /grill-me followed by your plan or idea.
Conclusion
The Grill Me Skill proves that prompt engineering isn't about complexity—it's about precision. Six carefully chosen lines create a structured interview process that systematically eliminates ambiguity and uncovers hidden requirements. In an era where AI-assisted development is becoming standard, tools like this demonstrate how simple prompts, when thoughtfully designed, can have outsized impact on software quality.
