facts-extraction-normalization

What problem does it solve?

This Skill formalizes extracting, normalizing, and managing longitudinal patient facts (meds, goals, incidents, services, providers, baselines) from current visit evidence and approved outputs, producing a reviewable facts table used as the primary guardrail for RAG and change detection. It enables auditability and deterministic comparisons across visits while discouraging memory-based hallucinations.

Core Features & Use Cases

• Extract candidate facts from today’s transcript segments and OCR regions
• Normalize and deduplicate facts for a consistent longitudinal view
• Propose updates (insert/update/resolve) for human review and controlled write-back
• Persist approved updates to the patient facts table and expose active facts to the RAG context pack
• Use cases include implementing fact schemas, controlled vocabularies, and write-back workflows in clinical data pipelines

Quick Start

Use the facts-extraction-normalization skill to generate candidate facts from today’s transcript and OCR outputs and prepare them for review.