A /grill-me Claude Skill Turns Out to Be a Sharp Technical Interview Tool

A skill built to interrogate your plan before you code doubles as a way to test how a candidate actually thinks with AI. Matt Pocock, the developer behind AI Hero, shares an unexpected use case someone sent him: running his /grill-me skill inside a live technical interview. The candidate joins a video call, opens Claude Code, and uses /grill-me while adding a feature to a codebase. What the skill demands from the user is exactly what the interviewer wants to observe.

The Use Case Someone Sent Him

Pocock opens by describing a message from a user who applied /grill-me in a way he had not considered. They were using it for technical interviews. The setup is simple: interview a candidate over a video call, have them open Claude Code, and get them to use /grill-me while working through adding a feature to a codebase. The point is not to watch the candidate produce clean output. The point is to watch how they respond when the tooling pushes back.

Why grill-me Works as an Interview

The /grill-me skill asks a lot of the user. As Pocock puts it, it forces you to clearly describe where you are going. It grills you with questions to test whether you have the right instincts and whether the thing you are building actually needs to exist. That interrogation is the mechanism. When a candidate has to defend a plan against a stream of pointed questions, the interviewer sees their reasoning, their assumptions, and whether they can articulate a direction under pressure.

The skill lives in Pocock's public mattpocock/skills collection and runs a structured session that challenges each branch of a proposed change. In practice it asks many questions per session, which is what makes it a stress test rather than a checklist.

What It Actually Measures

Pocock frames the qualities the skill surfaces. You need a long-term mindset. You need to clearly describe what you want. Those are the strategic instincts that matter when working with AI, not the mechanical ability to type prompts. Because the candidate has to steer the skill and answer its questions, the interviewer can assess how the person thinks with AI, how they work with AI, and whether they think strategically about what they are building.

The Design Philosophy Behind It

This use case fits how Pocock wants his skills to be used. He says he designs them to empower the user while demanding a lot of skill from the user to work best. That is a deliberate stance. A tool that carries all the weight tells you nothing about the person holding it. A tool that requires clear thinking, direction, and judgment reveals whether those things are present. The same property that makes /grill-me useful for planning real work makes it a fair signal in an interview.

Simple Enough That Anyone Can Pick It Up

Pocock closes on accessibility. Alongside its demands, /grill-me is simple enough that anyone can pick it up. That combination matters for an interview: the candidate does not need prior familiarity with the specific skill to use it, so the exercise tests thinking rather than tool memorization. And, as he notes, it is handy for anyone who wants to get better at working with AI, interview or not.

Key Takeaways

/grill-me was designed to interrogate a plan before coding starts, but it also works as a live technical interview tool.
The interview format: candidate on a video call opens Claude Code and uses /grill-me while adding a feature to a codebase.
The skill demands clear description of direction and defends against pointed questions, which surfaces how a candidate reasons.
It measures strategic qualities: long-term mindset, clear articulation of intent, and how a person thinks and works with AI.
Pocock intentionally designs skills that empower users while demanding skill from them, which is why the tool doubles as a signal.
The skill is simple enough to pick up cold, so the exercise tests thinking rather than familiarity with the tool.