Clinical Decision-Support Research & Evaluation
Drafts and validates research-only CDS artifacts — intended-use statements, GRADE evidence profiles, aggregate cohort tables, survival plans, model evaluations, and privacy/governance checklists.
What it does
Produces and checks documentation and evaluation artifacts for clinical decision-support research using bundled templates and offline scripts.
- Intended-use / limitation statements plus a mandatory artifact header (version, owner, approval boundary, prohibited uses)
- GRADE evidence-profile completeness and citation-link checking (humans still assign certainty)
- Aggregate cohort tables with minimum-cell thresholds and primary/complementary suppression
- Survival-analysis plan validation: time zero, competing events, censoring, estimand, proportional-hazards checks
- Aggregate model/biomarker evaluation from confusion counts: Wilson intervals, calibration gaps, subgroup differences, explicit suppression
- De-identification process checklist and decision-logic traceability matrix
Every script is standard-library only — no network, no API keys, no LLM calls. Patient rows, identifiers, free text, images, waveforms, and genomic sequences are rejected at the data gate.
Note the skill's own hard boundary: never for diagnosis, treatment decisions, dosing, triage, alerts, or live clinical operation.
Who it's for
- Clinical researchers and biostatisticians standardizing SAPs and reporting-guideline compliance
- Health-AI teams preparing TRIPOD+AI, PROBAST+AI, STARD-AI, or CONSORT-AI documentation
- IRB and data-governance staff formalizing de-identification and disclosure-control policy
- RA/QA writers drafting regulatory context (FDA CDS guidance, ONC HTI-1, ICH E6(R3)/E9)
Examples
- Model performance report — feed aggregate confusion counts from an external validation cohort; get sensitivity/specificity with Wilson intervals and automatic suppression of thin subgroup cells.
- Evidence-profile audit —
evidence_profile_check.pyflags GRADE rows missing risk-of-bias, inconsistency judgments, or source IDs before panel sign-off. - Survival plan pre-check — confirm time origin, immortal-time risk, competing-risk handling, and an RMST fallback are all specified before the protocol is locked.
· · · Install guide · · ·
Install in the Claude app (no terminal)
- Download the ZIP with the button below.
- In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
- Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
Install in Claude Code
Let Claude do it — paste this into Claude Code
Install the skill I found on Claude Skill Mart. Copy the skills/clinical-decision-support folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/clinical-decision-support/. When it's done, tell me in one line what this skill can do.
Install with a command instead
git clone https://github.com/K-Dense-AI/scientific-agent-skills.git && mkdir -p ~/.claude/skills && cp -r scientific-agent-skills/skills/clinical-decision-support ~/.claude/skills/⚠ This is a third-party skill. Check the source repository before installing.
- Open a terminal and verify Python:
python3 --version(3.11+ required). - Clone the repository:
git clone https://github.com/K-Dense-AI/scientific-agent-skills.git - Create the skills directory:
mkdir -p ~/.claude/skills - Copy just this skill:
cp -r scientific-agent-skills/skills/clinical-decision-support ~/.claude/skills/ - Restart Claude Code, then trigger it with something like "Use the clinical-decision-support skill to build an aggregate cohort table."
- Optional smoke test from the skill directory:
python3 scripts/validate_cds_artifact.py --help - Only supply synthetic or aggregate inputs — never patient-level rows or identifiers.