Open the idea up before you narrow it down.
Paste one raw idea - a product, a feature, a tool, a new process. Your browser checks it against the idea-refine skill's five sharpening questions - who it is for, what success looks like, the constraints, what has been tried, why now - and its red flags, for free. Expand reframes it as a How-Might-We question, asks what it still needs to know and writes one variation per ideation lens. Tick the ones that resonate, answer the questions, and Converge clusters them into two or three directions, rates each, and writes a one-pager with the assumptions to test, the MVP and a Not Doing list.
The ideas, schools and products are invented. Each example comes with a saved model reply, so you can see the whole page without signing in or spending a credit.
What this does, and what it does not
Most ideas die of two opposite mistakes: building the first version you thought of, or never choosing. The idea-refine skill runs them in order - open the idea up (divergent), then evaluate and converge, then ship a one-pager - and insists on two things people skip: who this is for, and the assumptions you are betting on. This page runs that method on one idea.
The free part runs in your browser: the five sharpening questions checked against your text, the red flags of the skill's refinement rubric ("everyone could use this", "like X but better", a technology before a person, too many features, no stated pain), a How-Might-We draft, and the whole method as a worksheet. Expand reframes the idea and writes one variation per lens - simpler, the opposite, another user, an expert's view, no limits, merged with an adjacent idea, massive scale. Converge clusters the variations you picked into directions, rates each as a painkiller or a vitamin, by feasibility and by kind of differentiation, recommends one with the skill's decision matrix, and writes the one-pager: problem statement, recommended direction, assumptions with a way to test each, the MVP, the Not Doing list and the open questions.
It never states market sizes, prices or user counts your input does not contain - those are assumptions to test, and the page checks for them. It does not validate the idea for you: the assumptions list is where that work starts. Derived from the agent skill @addyosmani/idea-refine (github.com/addyosmani/agent-skills, MIT). Sources are listed in llms.txt.