What problem does it solve? When a person describes a problem with their EPS (Colombian health insurer) in their own words, someone must reliably pull out the facts needed to route the case, without guessing, inferring, or making legal decisions. This Skill performs that extraction with strict evidence requirements so downstream deterministic logic can decide the legal route safely. ## Core Features & Use Cases - Evidence-bound extraction: Every extracted field reports a value, a confidence score (0.0-1.0), and a verbatim quote from the user as evidence; if no literal quote supports a value, it must be null. - Four triage slots plus case data: Extracts riesgo_vital, sujeto_especial, urgencia, tutela_previa_cumplida, solicitud_previa (ninguna/verbal/escrita), termino_vencido, plus identity fields, EPS, denied service, and desacato indicators. - Patient role detection: Determines whether the case concerns the speaker ("yo"), a minor ("menor"), or another adult ("otro"), which selects the correct legal document template. - Use Case: A user says "llevo tres semanas yendo a la farmacia por la insulina y me dicen que vuelva mañana". The Skill reports solicitud_previa as "verbal" with the literal quote, avoiding the costly error of marking it "ninguna" and sending the person down the wrong legal path. ## Quick Start Give the Skill a transcript of what a person said about their EPS problem and ask it to extract the triage slots and case data with evidence quotes.
