ysat-jevlisted

## Overview This is the typed judgment variant of the `ysat` skill (You Sure About That?). Everything about the purpose is identical: it disagrees with a decision, with evidence, before the decision is made. The difference is in one layer, and only one. In `ysat`, the model reads the context and writes the verdict in prose. In `ysat-jev`, the model never writes a verdict at all. It answers **atomic typed questions** about the decision, one at a time, and the **verdict is composed in code** from those answers with coefficients you can read and change. That inversion buys the thing prose cannot give: the verdict is a deterministic function of the state. The same state produces the same verdict, every run, on any day, no matter how the user phrased the request or how hard they pushed back. Rhetorical pressure is not a field in the state, so it has no path to the result. **When in doubt between the two: use `ysat`.** This variant costs more steps and returns a colder artifact. It earns that cost when the verdict has to be defended, replayed or audited later. ## When to Use - The decision will be contested and the reasoning has to survive being re-read weeks later. - The user wants the same decision judged more than once, and the runs must be comparable. - The user explicitly asks for a typed verdict, an audit trail, or `modo jev`. - A previous counterpoint softened under pushback and the user wants that made impossible. ## When NOT to Use - A quick counterpoint in convers