s6ig1t

Brightspring Referral Agent — Data skill for Claude Code

Data community

An agentic pipeline that reads a home-health, hospice, or behavioral-health referral document, checks it against deterministic clinical rules, and drafts a cited review brief for a pharmacist or intak.

How to install Brightspring Referral Agent

This entry records only its repository, not the path inside it, so there is no exact command to give. Open s6ig1t/brightspring-referral-agent and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Brightspring Referral Agent does

An agentic pipeline that reads a home-health, hospice, or behavioral-health referral document, checks it against deterministic clinical rules, and drafts a cited review brief for a pharmacist or intake nurse, where every finding traces back to the exact page and snippet it came from, and nothing is finalized without a named human's approval.

Alternatives in Data

  • Data Contract — This document defines which files belong to the system (auto-updatable) and which belong to the user (never to 32.5k ★
  • Clinical Decision Support — Evidence-based clinical decision support tools 6.4k ★
  • L Docs Architecture — Document cross-cutting Lowdefy architecture topics (build pipeline, request lifecycle, etc 3k ★

README

BrightSpring Referral Intake & Review Pipeline

An agentic pipeline that reads a home-health, hospice, or behavioral-health referral document, checks it against deterministic clinical rules, and drafts a cited review brief for a pharmacist or intake nurse, where every finding traces back to the exact page and snippet it came from, and nothing is finalized without a named human's approval.

Built as a hands-on portfolio project applying agentic AI development- the Microsoft Agent Framework (`Microsoft.Agents.AI`), Claude, and .NET 10- to a realistic home- and community-based healthcare workflow, in the style of home health, hospice, and pharmacy coordination services that companies like BrightSpring Health Services operate at scale.

**This project is independent, unaffiliated portfolio work.** It is not built with, endorsed by, or reviewed by BrightSpring Health Services or any other company. All referral documents in `samples/` are entirely synthetic- invented names, invented conditions, invented medication lists; and the drug reference data in `BrightSpring.Referral.Rules/DrugReferenceData.cs` is a small, self-authored table for demonstration purposes, **not clinically validated data**. See that file's header comment for what a production version would replace it with.

The core design rule

**The model reads the document and writes the narrative. It never touches the clinical judgment.**

Every finding- a duplicate drug class, a dose above reference range, an allergy conflict, a high-risk combination- is computed by plain, deterministic, unit-tested C#. The LLM is never asked to decide whether something is safe, and is never shown the raw document when drafting the review narrative; it only ever sees already-computed, already-correct findings. That separation isn't just a design intention- it's enforced by the project structure itself: `BrightSpring.Referral.Rules` has zero dependency on anything AI-related, and could not call into an LLM even if a future co