| name | clinical-decision-support-review |
| description | Review clinical decision support, treatment-plan, and clinical-report workflows for evidence grounding, patient-safety language, scope limits, and no-PHI handling. Use when a task asks for CDS, care pathways, treatment plans, or clinical report drafting. |
| user-invocable | true |
| license | Apache-2.0 |
Clinical Decision Support Review
Source adaptation: BioRouter-original synthesis inspired by remaining Scientific Agent Skills and Auto-Empirical Research Skills capability clusters, rewritten under Apache-2.0. This is not copied upstream text; it is a native BioRouter routing and verification checklist.
When to Use
Use this skill when the user asks about clinical decision support, care pathways, treatment plans, clinical reports, safety claims, or no-PHI review. Prefer the most specific BioRouter extension, BioSkill, or language style skill whenever a concrete platform, package, or file type appears.
Intake
Before producing output, capture:
- Research or operational question.
- Data, platform, package, or source material involved.
- Unit of analysis, model system, cohort, sample, or artifact.
- Desired deliverable: plan, code, review, table, figure, protocol, or narrative.
- Constraints: credentials, privacy, licensing, compute budget, validation standard, or external service availability.
BioRouter Routes
Use these routes when applicable:
- Platform/API work: call the relevant BRXT extension first and keep mutation/execution gated.
- Biomedical or omics data: load the relevant
bio-skills category.
- Statistical/causal claims: load
causal-identification-gates, statistical-modeling-bayes, or econometrics-toolkit.
- Evidence/citations: load
citation-temporal-integrity and claim-evidence-integrity.
- Reproducibility: load
replication-package-audit before treating the result as final.
Quality Gates
Do not present final claims until these are true:
- Provenance is explicit: source files, APIs, ids, versions, dates, or citations.
- The method matches the data-generating process and the intended claim.
- Assumptions and failure modes are listed plainly.
- Validation is tied to an actual benchmark, holdout, sensitivity check, audit trail, or source record.
- The output can be rerun, reviewed, or traced by another person.
- Anything requiring external credentials, proprietary data, wet-lab execution, or live services is marked unverified unless actually checked.
Response Shape
Return:
- Recommended route.
- Minimal next action.
- Verification checklist.
- Risks or blockers.
- Exact artifact to produce next, if any.
