What problem does it solve? Turning natural-language understanding into reliable application behavior usually means fragile LLM prompt-and-parse steps. This Skill guides you to replace those with TypeSafe's System One models (like Jev), which return typed judgments and probabilities that ordinary code can combine, route on, and verify. ## Core Features & Use Cases - Typed judgment primitives: Design Choice, Noul, and Score questions over structured state so code receives typed answers and probabilities instead of generated text. - Composition patterns: Apply proven shapes such as routing with argument filling, select-instead-of-generate extraction, reranking, composite scoring, verification with escalation, and state-responsive agents. - Live documentation workflow: Read the current TypeSafe docs, API references, SDK pages, and cookbooks as the source of truth before writing any integration. - Use Case: When building a support ticket router, ask branch-specific questions up front, consume only the relevant typed answers, and let code own the workflow while the model supplies the semantic judgment. ## Quick Start Ask the AI to design a TypeSafe integration that routes incoming user requests to typed handlers using Jev judgments instead of an LLM prompt-and-parse step.
