/grill-me is a skill that pushes a plan or idea forward by asking relentless, structured questions.
In simple terms, it helps Claude challenge vague thinking before the real work starts. Instead of letting an unclear request slide through, it keeps asking until the goal, constraints, and missing assumptions become visible.
That is why it works especially well right after opening Claude. At that moment, the real bottleneck is often not the model. It is the human side: unclear goals, half-formed plans, and unstated assumptions.
When I use /grill-me early, the rest of the session usually goes more smoothly. It feels slower for a few minutes, but it reduces the number of times I have to stop, rethink the scope, or correct the direction later.
The skill is also practical because it does not blindly ask everything back to the user. If something can be answered by checking the codebase or files, Claude is supposed to inspect those first and only ask where necessary.
A simple way to use it is to start with a short brief like this:
/grill-me
I am planning a blog post about this topic.
The reader is new to Claude skills.
The goal is to explain the value of /grill-me clearly.
Ask me the missing questions one by one, and give a recommended answer each time.My impression is that /grill-me is not a friendly brainstorming helper as much as it is a precision tool. It improves the quality of the human request. That matters because AI output often reflects the quality of the input more than people want to admit.
If your sessions often drift because the plan was still fuzzy, /grill-me is worth trying at the beginning.
