agentleFS
NVIDIAKnown publisherNVIDIA/dgx-spark-playbooks/nvidia/station-healthcare-agent/assets/skills/clinical-knowledge/SKILL.md

clinical-knowledge

Skill1.5k starsChanged 2 months agoApache-2.0View on GitHub

Teaches agents clinical reference ranges, condition codes, quality measure definitions, drug classifications, and regulatory context so they can flag abnormal values and identify care gaps.

Trending

More talk lately.about the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 0 mentionsMar 2026: 0 mentionsApr 2026: 0 mentionsMay 2026: 0 mentionsJun 2026: 0 mentionsJul 2026: 0 mentionsAug 2026: 0 mentionsSep 2026: 0 mentionsOct 2026: 48 mentionsNov 2026: 2 mentions
Dec 2025Jun 2026Nov 2026
Mentions a month of the NVIDIA/dgx-spark-playbooks repository that our weekly crawl found. There is no adoption data for this file yet.

Who's talking about NVIDIA/dgx-spark-playbooks

The biggest 5 of 50 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 50

Talked about on GitHub (50 times).

On GitHub · 50

45 more on GitHub

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When reporting lab values:

    line 42
  2. Distinctive:

    Always flag values outside the normal range with the severity (mild / moderate / severe)

    line 43
  3. Very specific:

    If the FHIR Observation includes a referenceRange, use that instead of the table above

    line 45
  4. Very specific:

    When FHIR Condition resources use ICD-10 coding (system http://hl7.org/fhir/sid/icd-10-cm), map as follows:

    line 66
  5. Very specific:

    When checking medication coverage, recognize these drug class groupings.

    line 116
  6. Distinctive:

    Never state that a patient "has" a condition based on a lab value alone

    line 193
  7. Distinctive:

    Always include the disclaimer: "This is for informational and research purposes, not clinical decision-making"

    line 194
  8. Distinctive:

    When flagging care gaps, use language like "may warrant review" not "requires treatment"

    line 195

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Clinical Reference Knowledge

Regulatory Context

The 21st Century Cures Act (2016) and ONC's Interoperability Final Rule (2020) require US healthcare organizations to expose patient data through standardized FHIR APIs. As of 2024, ~70% of US hospitals support FHIR R4 (source: ONC). CMS ties quality measure reporting to reimbursement through programs like MIPS (Merit-based Incentive Payment System) and the Hospital Value-Based Purchasing Program. Failure to report -- or poor performance -- results in payment adjustments of up to 9%.

HIPAA (Health Insurance Portability and Accountability Act) prohibits transmitting Protected Health Information (PHI) to external services without a BAA (Business Associate Agreement). This is the primary reason clinical AI must run locally: cloud LLM APIs are not BAA-covered by default, and even those that offer BAAs (e.g., Azure OpenAI) face institutional resistance from hospital compliance teams.

Common Lab Reference Ranges

These are general adult reference ranges. Values vary by lab, assay, and patient characteristics. Always defer to the performing laboratory's reference range when available via FHIR referenceRange.

LabNormal RangeConcerningUnitNotes
HbA1c< 5.7% (non-diabetic), < 7.0% (diabetic target)> 9.0% = poor control%ADA 2024 guidelines; target may be relaxed to < 8.0% for elderly/frail
Fasting Glucose70-100100-125 = prediabetes, >= 126 = diabeticmg/dLMust be fasting; random glucose >= 200 also diagnostic
Creatinine0.7-1.3 (male), 0.6-1.1 (female)> 1.5 = impaired renalmg/dLAffected by muscle mass; less reliable in elderly
eGFR> 90 (normal), 60-89 (mild decrease)30-59 = moderate CKD, 15-29 = severe, < 15 = kidney failuremL/min/1.73m2CKD-EPI 2021 equation (race-neutral)
BUN7-20> 20 with rising creatinine = renal concernmg/dLElevated by dehydration, high protein diet
Total Cholesterol< 200200-239 = borderline, >= 240 = highmg/dL
LDL< 100 (general), < 70 (high-risk ASCVD)> 160 = highmg/dLACC/AHA 2018
HDL> 40 (male), > 50 (female)< 40 = cardiovascular risk factormg/dL
Triglycerides< 150150-499 = elevated, >= 500 = severe (pancreatitis risk)mg/dL
Systolic BP< 120 (normal), 120-129 (elevated)130-139 = Stage 1 HTN, >= 140 = Stage 2 HTNmmHgACC/AHA 2017; CMS165 uses 140/90 threshold
Diastolic BP< 8080-89 = Stage 1 HTN, >= 90 = Stage 2 HTNmmHg
BNP< 100100-400 = possible HF, > 400 = likely HFpg/mLAge-adjusted: higher cutoffs in elderly; obesity lowers BNP
NT-proBNP< 300 (rule-out)Age-stratified: >450 (<50y), >900 (50-75y), >1800 (>75y)pg/mLMore stable than BNP; renal clearance affects levels
Potassium3.5-5.0< 3.5 = hypokalemia, > 5.5 = hyperkalemia (cardiac risk)mEq/LCritical for patients on ACEi/ARB/spironolactone
Sodium136-145< 130 = moderate hyponatremiamEq/LCommon in HF patients
ALT7-56> 3x ULN = significant hepatotoxicityU/LMonitor with statin therapy
Hemoglobin13.5-17.5 (male), 12.0-16.0 (female)< 12 (male) or < 11 (female) = anemiag/dLCommon in CKD (erythropoietin deficiency)

When reporting lab values:

  • Always flag values outside the normal range with the severity (mild / moderate / severe)
  • Note the date of the observation -- a result from 2 years ago has different clinical significance than one from yesterday
  • If the FHIR Observation includes a referenceRange, use that instead of the table above

Condition Codes

SNOMED CT Codes (Primary)

CodeConditionICD-10 EquivalentPrevalence (US adults)
44054006Type 2 Diabetes MellitusE11.x~11% (37M)
46635009Type 1 Diabetes MellitusE10.x~0.5% (1.6M)
38341003Essential HypertensionI10~47% (116M)
84114007Heart FailureI50.x~2.4% (6.7M)
40055000Chronic Kidney DiseaseN18.x~15% (37M)
53741008Coronary Artery DiseaseI25.x~7% (20M)
13645005COPDJ44.x~6% (16M)
195967001AsthmaJ45.x~8% (25M)
49436004Atrial FibrillationI48.x~2% (6M)
73211009Diabetes (unspecified)E11.9Used in older records

ICD-10-CM to SNOMED Crosswalk

When FHIR Condition resources use ICD-10 coding (system http://hl7.org/fhir/sid/icd-10-cm), map as follows:

  • E11.* → Type 2 Diabetes (SNOMED 44054006)
  • E10.* → Type 1 Diabetes (SNOMED 46635009)
  • I10 → Essential Hypertension (SNOMED 38341003)
  • I50.* → Heart Failure (SNOMED 84114007)
  • N18.* → Chronic Kidney Disease (SNOMED 40055000)

Note: A Condition resource may have both SNOMED and ICD-10 codes in the coding array. Always check all entries, not just coding[0].

CMS Quality Measures

CMS122v12 -- Diabetes: Hemoglobin A1c (HbA1c) Poor Control (> 9%)

ComponentDefinition
DenominatorPatients 18-75 with diabetes (Type 1 or Type 2) and at least 2 encounters during the measurement period
NumeratorPatients with most recent HbA1c > 9.0%, OR no HbA1c recorded during the measurement period
ExclusionsHospice care, palliative care, advanced illness with frailty (2+ encounters for advanced illness AND frailty diagnosis), dementia medications (donepezil, rivastigmine, memantine, galantamine)
Performance rateLower is better (inverse measure)
Payment impactPart of MIPS quality reporting; affects Medicare reimbursement

CMS165v12 -- Controlling High Blood Pressure

ComponentDefinition
DenominatorPatients 18-85 with essential hypertension diagnosed before or during the measurement period
NumeratorPatients with most recent BP < 140/90 mmHg
ExclusionsHospice, palliative care, ESRD, kidney transplant, advanced illness with frailty, pregnancy
Performance rateHigher is better
NoteBP must be measured during an outpatient encounter; home BP readings are not counted in the standard measure

CMS135v12 -- Heart Failure: ACEi/ARB/ARNI Therapy for LVEF < 40%

ComponentDefinition
DenominatorPatients 18+ with heart failure AND documented LVEF < 40% (HFrEF)
NumeratorPatients prescribed ACE inhibitor, ARB, or ARNI (sacubitril/valsartan)
ExclusionsHospice, allergy/intolerance to all three classes, bilateral renal artery stenosis, pregnancy, hyperkalemia > 5.5
NoteLVEF data often in DiagnosticReport or CarePlan, not always queryable via Condition alone

CMS134v12 -- Diabetes: Medical Attention for Nephropathy

ComponentDefinition
DenominatorPatients 18-75 with diabetes
NumeratorPatients with nephropathy screening (urine albumin test) OR evidence of nephropathy treatment (ACEi/ARB) OR nephropathy diagnosis
ExclusionsHospice, palliative care, advanced illness with frailty

Drug Classifications

When checking medication coverage, recognize these drug class groupings. Matching should be case-insensitive partial string matching on the medication name from FHIR medicationCodeableConcept.text or .coding[].display.

Diabetes Medications

ClassDrugsNotes
BiguanidemetforminFirst-line therapy
Sulfonylureasglipizide, glyburide, glimepirideHypoglycemia risk
Insulininsulin lispro, insulin glargine, insulin aspart, insulin detemir, insulin degludec, NPH insulinMatch any string containing "insulin"
GLP-1 Receptor Agonistsliraglutide (Victoza), semaglutide (Ozempic/Wegovy/Rybelsus), dulaglutide (Trulicity), exenatide (Byetta/Bydureon), tirzepatide (Mounjaro)Weight loss benefit; cardiovascular benefit
SGLT2 Inhibitorsempagliflozin (Jardiance), dapagliflozin (Farxiga), canagliflozin (Invokana), ertugliflozin (Steglatro)Cardiovascular + renal benefit; monitor for DKA
DPP-4 Inhibitorssitagliptin (Januvia), saxagliptin, linagliptin, alogliptinWeight-neutral
Thiazolidinedionespioglitazone, rosiglitazoneHF risk; edema

Antihypertensives

ClassDrugsNotes
ACE Inhibitorslisinopril, enalapril, ramipril, benazepril, fosinopril, quinaprilCough side effect; monitor K+ and creatinine
ARBslosartan, valsartan, irbesartan, olmesartan, telmisartan, candesartan, azilsartanAlternative if ACEi cough
ARNIssacubitril/valsartan (Entresto)HFrEF guideline-directed; do NOT combine with ACEi
CCBsamlodipine, nifedipine, diltiazem, verapamilDiltiazem/verapamil contraindicated in HFrEF
Beta-Blockersmetoprolol (tartrate or succinate), atenolol, carvedilol, bisoprolol, propranolol, nebivololOnly carvedilol, metoprolol succinate, bisoprolol for HF
Thiazide Diureticshydrochlorothiazide (HCTZ), chlorthalidone, indapamideFirst-line for HTN
Loop Diureticsfurosemide, bumetanide, torsemideVolume management in HF, not primary HTN therapy
Aldosterone Antagonistsspironolactone, eplerenoneHF benefit; monitor K+

Statins (HMG-CoA Reductase Inhibitors)

IntensityDrugs
Highatorvastatin 40-80mg, rosuvastatin 20-40mg
Moderateatorvastatin 10-20mg, rosuvastatin 5-10mg, simvastatin 20-40mg, pravastatin 40-80mg
Lowsimvastatin 10mg, pravastatin 10-20mg, lovastatin 20mg

Heart Failure Medications (Guideline-Directed Medical Therapy)

The four pillars of HFrEF therapy (ACC/AHA 2022):

  1. ACEi/ARB/ARNI -- sacubitril/valsartan preferred over ACEi/ARB
  2. Beta-Blocker -- carvedilol, metoprolol succinate, or bisoprolol only
  3. Aldosterone Antagonist -- spironolactone or eplerenone (if eGFR > 30, K+ < 5.0)
  4. SGLT2 Inhibitor -- dapagliflozin or empagliflozin (regardless of diabetes status)

Matching Strategy

medication_name = fhir_med_text.lower()

is_on_insulin = "insulin" in medication_name
is_on_glp1 = any(drug in medication_name for drug in
    ["liraglutide", "semaglutide", "dulaglutide", "exenatide", "tirzepatide",
     "victoza", "ozempic", "trulicity", "byetta", "mounjaro", "rybelsus"])
is_on_sglt2 = any(drug in medication_name for drug in
    ["empagliflozin", "dapagliflozin", "canagliflozin", "ertugliflozin",
     "jardiance", "farxiga", "invokana", "steglatro"])
is_on_acei = any(drug in medication_name for drug in
    ["lisinopril", "enalapril", "ramipril", "benazepril", "fosinopril", "quinapril"])
is_on_arb = any(drug in medication_name for drug in
    ["losartan", "valsartan", "irbesartan", "olmesartan", "telmisartan",
     "candesartan", "azilsartan"])
is_on_betablocker = any(drug in medication_name for drug in
    ["metoprolol", "atenolol", "carvedilol", "bisoprolol", "propranolol", "nebivolol"])
is_on_statin = any(drug in medication_name for drug in
    ["atorvastatin", "rosuvastatin", "simvastatin", "pravastatin", "lovastatin"])

Clinical Comorbidity Patterns

Common co-occurring conditions to watch for during analysis:

  • Cardiorenal-metabolic overlap: Diabetes + Hypertension + CKD occur together in ~30% of diabetic patients
  • Heart failure + CKD: eGFR < 30 limits medication options (spironolactone, SGLT2i dose adjustment)
  • Diabetes + CAD: Statin therapy should be high-intensity; GLP-1/SGLT2i have cardiovascular benefit
  • Atrial fibrillation + Heart failure: Common pairing; rate control with beta-blocker preferred

Guardrails

  • These reference ranges are general guidelines based on published clinical guidelines (ADA, ACC/AHA, KDIGO), not diagnostic criteria
  • Never state that a patient "has" a condition based on a lab value alone
  • Always include the disclaimer: "This is for informational and research purposes, not clinical decision-making"
  • When flagging care gaps, use language like "may warrant review" not "requires treatment"
  • Acknowledge that CMS measure logic here is simplified -- production implementations use the full eCQM (electronic Clinical Quality Measure) specifications from CMS
  • Lab reference ranges should defer to the performing laboratory's range when available

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.