Prompt Library¶
Ready-made prompts for studying, teaching, and research: copy one and fill in its [BRACKETED] parts, or open it to see how to use it.
Color key: Prompt
All 24 prompts
Research
Literature Screening Assistant
Second-screens titles and abstracts against your stated inclusion criteria
Critical Appraisal Second Reader
Checks your own appraisal of a paper, item by item, against its text
Reporting Guideline Auditor
Maps a manuscript against its reporting guideline, item by item
Further reading (2)
- Performance of a large language model in screening citations : A prospective test of model-assisted title and abstract screening for guideline development. Note how much revising the prompt's inclusion criteria mattered, with sensitivity rising from 0.75 to 0.91.
- Development and evaluation of prompts to screen titles and abstracts : Walks through the prompt-refinement workflow itself (a development set, iteration, then held-out evaluation) for literature screening in a living systematic review.
Multiple-Choice Question (MCQ) Writing
NBME-Style Question Tutor
Drills National Board of Medical Examiners (NBME) style questions built from your own course materials
The tutor builds its questions from your own materials, which suits your exams best, since in-house exams are written from them. Question-bank and NBME items are licensed content whose license may not allow this use.
Single Best Answer Item Writer
Drafts exam-quality single best answer items from your learning objectives
Team-Based Learning Session Builder
Builds a team-based learning session: scratch-card readiness tests and application cases
Further reading (2)
- Validity of AI-generated multiple-choice questions in medical education: a systematic review : A systematic review of 71 studies. Models proved useful drafting tools, error rates ranged from under 1% to 45%, and the evidence does not yet support using their items on a summative exam without review.
- GPT-4 versus human authors in clinically complex MCQ creation : A blinded panel compared items written by GPT-4 with human-written items at novice and expert level. Overall quality matched the expert items, but more of the model's items had a wrong key or needed major revision.
MCQ Review
Post-exam Item Analysis Reader
Flags post-exam item statistics for exam committee review, from numbers only
Further reading (3)
- Assessing multiple-choice question quality in internal medicine: three large language models against expert consensus : Three models and three experts reviewed the same 85 clerkship items. The stronger models matched the experts on objective flaws such as negatively worded stems, and fell short on judgment calls and on fit to the learning outcome.
- Faculty versus artificial intelligence chatbot: MCQ quality in physiology : Students sat 50 faculty-written and 50 chatbot-written physiology items. The chatbot's items discriminated about as well but were easier and had more distractors almost no one chose; its measures are the ones a post-exam item analysis reports.
- Application of artificial intelligence generated content in medical examinations : Reviews how generated questions fit into examination workflows, including the points where human review cannot be skipped.
Data Analysis
Content Generation
Learning Objective Sharpener
Tightens draft learning objectives so each can be taught and assessed
Lecture Outline Builder
Turns objectives into a session outline with active-learning checks
Feedback on Student Work
Rubric Feedback Drafter
Drafts quote-anchored observations on de-identified student writing against your frozen rubric
Study Strategy
NBME-Style Question Tutor
Drills National Board of Medical Examiners (NBME) style questions built from your own course materials
The tutor builds its questions from your own materials, which suits your exams best, since in-house exams are written from them. Question-bank and NBME items are licensed content whose license may not allow this use.
Score Report Study Planner
Turns an exam performance report into an honest, interleaved study plan
Study Plan Progress Check
Compares two reports and says honestly whether the plan is working
Study Schedule Builder
Builds a day-by-day schedule for a teaching block or a dated exam
Lecture Tutor
Teaches an attached lecture one objective at a time, checking you as it goes
Daily Review Sheet Builder
Builds a detail-preserving review sheet with self-test questions from one lecture
Missed-Question Debrief
Sorts the questions you missed by why you missed them, with a fix for each
Clinical Reasoning Partner
Practice differentials on invented cases or your own abstracted encounters, with reasoning feedback
Further reading (2)
- INSIGHTS interactive demo : The National Board of Medical Examiners' own clickable demo of the examinee dashboard, a quick way to see what your report will look like before your first exam.
- INSIGHTS user guide : The official walkthrough of every tab of the dashboard, including what the comparison flags mean, which reports carry per-question timing, and how to export your data.
Residency Applications
Mock Residency Interview
Rehearses residency interviews one question at a time, then debriefs from your answers
Personal Statement Critic
Critiques a personal statement you wrote, with questions instead of rewrites
Prompt Exchange¶
Share prompts that work for you and vote on what others have shared. Browse everything on the Prompt Exchange page, updated from the discussion board several times a day. Posting and voting need a free GitHub account; reading needs no account. Community prompts that hold up in testing can be promoted into the library above, with credit to the contributor.
No community prompts yet. Be the first to share one.
Prefer not to post publicly? Send prompts directly; contact information is on the About page.
About These Prompts¶
What the Policy Says about What You Paste¶
The prompts carry no data rules of their own; the Artificial Intelligence (AI) Responsible Use Policy is the reference for patient information, student records, and other sensitive material. In practice, a case or a piece of student work rarely needs names, ID numbers, dates, or places, and the prompts work just as well without them. The same goes for a detail unusual enough to recognize, such as a rare diagnosis on a small ward, since it can point to one person even without a name.
Two habits that make every prompt work better
Two habits make every prompt here work better: give the model your actual materials rather than a blank-page request, and ask for reasoning you can check rather than bare answers (Getting Better Answers covers the mechanics). New to prompting? Start with Learning to Prompt; related studies and guides appear as further reading under several sections above.
Draft and Reviewed: what review adds (you can use both)
Draft prompts are still being tested and refined, so their wording may still change. Reviewed prompts have completed that testing; that is what review adds. Neither label is a promise about what a model gives back, so check the output of either against your own materials and expertise. Like every review status in this library, the label is provisional until the AI Committee ratifies it.
What to expect across different models
These prompts are written for current assistants from Anthropic, OpenAI, and Google. The same prompt will not give identical results everywhere: output depends on the model's capability, the company that trained it, the material and context you provide, and the model version, which changes over time. Treat a weaker or older model's output as a rough draft, and judge every result against your own expertise. The structure and rules in each prompt carry across models even when the exact wording of the output does not.
Content last reviewed September 1, 2026. Review dates are maintained as described on the About page.