Agent Skills › NVIDIA/dgx-spark-playbooks › clinical-delegation

clinical-delegation

GitHub

指导如何委托临床任务给专家智能体。要求使用sub-agent运行时并指定agentId，严禁协调者直接调用FHIR API。涵盖患者数据、检验指标、药物及分析等场景的委派规范。

nvidia/station-healthcare-agent/assets/skills/clinical-delegation/SKILL.md NVIDIA/dgx-spark-playbooks

触发场景

需要获取患者数据或实验室结果执行临床数据分析或护理差距检查询问医疗代理委派规则

安装

npx skills add NVIDIA/dgx-spark-playbooks --skill clinical-delegation -g -y

更多选项

非标准路径

npx skills add https://github.com/NVIDIA/dgx-spark-playbooks/tree/main/nvidia/station-healthcare-agent/assets/skills/clinical-delegation -g -y

不安装直接使用

npx skills use NVIDIA/dgx-spark-playbooks@clinical-delegation

指定 Agent (Claude Code)

npx skills add NVIDIA/dgx-spark-playbooks --skill clinical-delegation -a claude-code -g -y

安装 repo 全部 skill

npx skills add NVIDIA/dgx-spark-playbooks --all -g -y

预览 repo 内 skill

npx skills add NVIDIA/dgx-spark-playbooks --list

SKILL.md

Frontmatter

{
    "name": "clinical-delegation",
    "metadata": {
        "openclaw": {
            "requires": {
                "bins": [
                    "python3"
                ]
            }
        }
    },
    "description": "How to delegate clinical tasks to specialist agents. Always use sub-agent runtime with explicit agentId — never ACP. Never call FHIR via web_fetch."
}

Clinical delegation (sessions_spawn)

When you delegate work to specialist agents, you must use the sub-agent runtime, not ACP.

Do not call FHIR yourself

Do not use web_fetch or any HTTP tool to call FHIR APIs from the coordinator. The coordinator does not have the correct FHIR base URL in context when it invents URLs. Specialist agents (patient-data, labs-vitals, medications, analyst) have the configured FHIR endpoint and write Python that runs on the server.

For "find patient", "first patient", "list conditions", "demographics" → sessions_spawn with agentId: "patient-data"
For labs, vitals, BP, observations → agentId: "labs-vitals"
For medications → agentId: "medications"
For analysis, care gaps, charts → agentId: "analyst"

The only FHIR server in use is https://r4.smarthealthit.org. Never use placeholder domains like fhir.example.com.

Rule

Use sessions_spawn with:

agentId — set to the specialist agent id: patient-data, labs-vitals, medications, analyst, or molecular
task — clear description of the task
Do not set runtime: "acp". Omit runtime or use runtime: "subagent" so the request goes to the correct clinical agent.

Correct examples

{ "agentId": "patient-data", "task": "Find the first patient and return demographics and active conditions.", "mode": "run" }

{ "agentId": "labs-vitals", "task": "Get latest HbA1c and eGFR for patient abc123", "mode": "run" }

{ "agentId": "analyst", "task": "Run care gap analysis for diabetic patients with A1c > 9%", "mode": "run" }

Wrong

Using web_fetch to call FHIR (e.g. https://fhir.example.com/Patient) — coordinator must delegate instead; fhir.example.com is not a real server and will fail.
"runtime": "acp" without agentId — causes "ACP target agent is not configured"
Omitting agentId when delegating — the system will not know which specialist to run

Specialist agents

agentId	Use for
patient-data	Find patients, get demographics, list conditions
labs-vitals	Labs, vitals, BP, observations
medications	Active medications, drug classes
analyst	Python analysis, care gaps, charts
molecular	Drug molecular structure, OpenFold3 NIM 3D visualization