wtf is clarify-crit?

Clarify, also called CRIT, is a small add-on skill meant to be loaded into an AI model such as Claude, designed to help the AI understand what someone actually wants before it starts working on their request. CRIT stands for Context, Role, Interview, Task, which describes the steps the skill guides the AI through: gathering what is already known, figuring out what role it should act in, asking only the specific questions that would actually change the outcome, and then carrying out the task. The idea behind the project is that many requests are ambiguous, and an AI that just guesses at missing details can produce work that misses the point. Rather than opening every single request with a long list of clarifying questions, Clarify tries to keep the back and forth to a minimum, asking just enough to remove real ambiguity, then either answering the request directly or handing back a short, reusable summary of what was clarified. The README gives a simple example: a user types Use CRIT, followed by a request like build a dashboard for my business, and the skill takes it from there. The repository is organized into two parts. One folder holds the actual skill package that gets uploaded into a Claude.ai skill uploader, or downloaded as a ready-made release file. The other folder holds testing and evaluation material used during development, which is not part of what actually runs when someone uses the skill. Full installation and usage instructions live in a separate README inside the skill folder itself. The project is released under the MIT license. Beyond this outline, the README is sparse and does not go into further detail about how the skill behaves internally.