wtf is build-or-skip?

Build or Skip is an AI workflow that evaluates a product idea and tells you honestly whether it is worth building. Most idea evaluation tools tend to encourage whatever you type in, but this one is designed to be willing to say no, and even to tell you to stop when an idea is not strong enough. The project is built on Dify, a platform for chaining AI steps together, rather than being a single prompt. The author explains that an earlier single prompt version was unreliable in specific ways: it almost always said BUILD even for ideas too large for one person, it sometimes changed its verdict without actually narrowing the scope of the idea, and mixing extraction, scoring, decision making, and report writing into one prompt made it hard to tell which part had failed. The current version splits the process into separate steps that each do one job. The workflow first extracts structured details from a rough idea, such as the target user, the core problem, and the main features. Next, an AI model scores the idea on three dimensions from one to five: how clear the target user is, how strong the underlying problem is, and how feasible a first version would be. A key design choice is that the actual verdict of BUILD, IMPROVE, or SKIP is decided by plain code with fixed rules, not by the AI model itself, so the same three scores always produce the same verdict. Depending on which verdict is reached, a different set of instructions generates the final report, and a SKIP verdict deliberately produces no build plan at all, only a short note that the problem is already solved well enough by existing options. The README walks through an example where the workflow initially approved a daily motivational quote app because it scored well on every measure, until the scoring rubric was adjusted to ask what a user does today without the product. That single change flipped the verdict to SKIP. To try it, you can use the hosted live demo, or import the included workflow file into Dify yourself and add your own OpenAI API key. It runs on the gpt-4o-mini model and a full evaluation costs roughly two tenths of a cent. The project is released under the MIT license.