The Anatomy of a Clinical Prompt
ROLE+
CONTEXT+
CONSTRAINTS+
TARGET OUTPUT=
Clinical Draft
Text in [brackets] is where you insert your specific patient data, clinical context, or details. Customize before use. When output is close but not right — iterate, don't start over.
Operating Rules
These six rules govern every prompt below. Each prompt inherits them; you don't have to restate them.
- 1You own the judgment. These prompts help you structure and check your own clinical reasoning. The AI organizes, drafts, and asks questions. It does not diagnose, decide, treat, or dose.
- 2Every AI output is a DRAFT for your review against your facility protocol and a licensed provider. It is never a finished clinical product.
- 3No PHI. Use a role, initials, or a bracketed placeholder. Never enter names, MRNs, full dates of birth, or anything that identifies a patient.
- 4No invented facts. The AI must not supply, calculate, or guess any value you did not give it — doses, thresholds, scores, lab ranges, or citations. If it doesn't have something, it must say so.
- 5Cite the current primary source. Require an organization, document title, and year (for example: AHA 2025; SSC 2026). If currency is uncertain, the AI must say so and you verify before use.
- 6Education and simulation only. Not medical, legal, or compliance advice. Not for clinical use.
A confident wrong answer is worse than a flagged uncertainty. If the AI can't support a clinical claim from a current authoritative source, it should say so rather than fill the gap.
No prompts match your search.