Source
When it runs
Only when you ask for it. The skill turns off automatic loading, so the agent never starts it on its own. This is the description it carries:
A relentless interview to sharpen a plan or design.
Why the skill is one line long
The whole body of grill-me is an instruction to call the grilling skill. The split is deliberate. Grilling holds the method and loads automatically when a conversation asks for it. Grill-me turns automatic loading off, so it is the version you type on purpose. Grill-with-docs is the same entry point with domain-modeling added, so the interview also writes decision records and a glossary as it goes.
What a grilling session looks like
The agent maps your plan as a design tree, where every decision branches into the decisions that depend on it. It asks in rounds. Each round covers the frontier: every open question whose prerequisites are already settled. Each question is numbered and comes with the agent's recommended answer, so you can agree in a word or argue.
Facts are the agent's job, not yours. When a question needs something from the codebase or the environment, the agent sends a sub-agent to look it up instead of asking you. The session ends when the frontier is empty and nothing is left silently assumed, and the agent does not act on the result until you confirm it.
What ships with it
Besides SKILL.md, the folder carries these. The agent opens them only when a step in the instructions calls for one.
-
agents/
Install the grill me skill
In Claude Code, run this in a session. It comes from the publisher's own instructions.
/plugin install mattpocock-skillsLicense: MIT.
In Longwave, add Engineering Skills from the marketplace. It installs all 37 of its skills, and your Claude Code, Codex and OpenCode threads can use them.