step1-notebooklm-guide

name	step1-notebooklm-guide
description	USMLE Step 1 domain knowledge for generating NotebookLM podcast prompts. Contains the proven prompt template, chapter-to-vignette mapping, high-yield topic patterns, and pedagogical strategies that produce 100% source coverage in NotebookLM's Audio Overview. Load this skill when generating prompts for any Survivor's Guide to USMLE Step 1 chapter.
argument-hint	[chapter number or name to generate a prompt for]

USMLE Step 1 NotebookLM Prompt Guide

This skill contains domain-specific knowledge for generating NotebookLM Audio Overview prompts for the Survivor's Guide to USMLE Step 1, 6th Edition. It encodes lessons learned from generating and testing prompts for Chapters 1 and 18, where coverage was measured by transcribing the generated podcast and comparing it against the source material.

Source Material Structure

The textbook has 23 chapters (00-22):

#	Chapter	Key Domain
00	Front Matter	—
01	General Principles	Autonomic pharmacology, PK/PD, drug reactions, CYP450
02	Shock and Its Management	Hemodynamics, shock types, fluid resuscitation
03	Endocrinology	Hormones, thyroid, adrenal, pituitary, diabetes
04	Reproductive System and OBGYN	Reproduction, pregnancy, contraception
05	Renal System	Nephron physiology, acid-base, electrolytes, diuretics
06	Gastroenterology	GI tract, liver, pancreas, biliary
07	Respiratory System	Lung mechanics, obstructive vs restrictive, gas exchange
08	Cardiology	Heart physiology, arrhythmias, valvular disease, HF
09	Genetics and Nutrition	Inheritance patterns, vitamins, minerals
10	Rheumatology and Dermatology	Autoimmune, connective tissue, skin
11	Neurology	CNS, PNS, cranial nerves, stroke, demyelinating
12	Psychiatry	Mood, psychotic, anxiety disorders, pharmacotherapy
13	Social Sciences and Ethics	Bioethics, epidemiology concepts
14	Biostatistics	Sensitivity, specificity, study design, bias
15	Infectious Disease	Bacteria, viruses, fungi, parasites, antibiotics
16	Biochemistry	Metabolic pathways, enzyme deficiencies, storage diseases
17	Hematology, Oncology and Immunology	Blood cells, clotting, cancers, immune system
18	Pathology	Cell injury, necrosis, inflammation, calcification
19	Poisoning and Miscellaneous Management	Toxicology, antidotes
20	Clinical Vignettes and Management	Pattern-recognition vignettes by system
21	Principles of Management	Imaging, procedures, cancer management
22	Test Taking Skills	Exam strategy

Supplementary Sources

Chapters 20, 21, and 22 are NOT standalone podcast chapters — they are supplementary material to enrich all other chapters:

• Chapter 20 (Clinical Vignettes): Short clinical scenarios with classic presentations → use to teach chapter concepts in Step 1 question stem language
• Chapter 21 (Principles of Management): When to order what test, imaging types, management protocols → use to add management context to chapter topics
• Chapter 22 (Test Taking Skills): Exam strategy → not used in prompts

Chapter-to-Vignette Mapping

When generating a prompt for chapter N, extract vignettes from chapters 20/21 that match chapter N's topics. Key mappings:

Chapter	Relevant Vignettes From Ch 20
01 General Principles	Parkinson's, Wilson's, Pheochromocytoma, MG, Delirium, Thyroid Storm
02 Shock	Cardiac tamponade, aortic dissection, MI presentations
03 Endocrinology	DKA, thyroid storm, Cushing's, Addison's, PCOS, pheochromocytoma
04 Reproductive	Abruptio placenta, placenta previa, ectopic, testicular torsion
05 Renal	Kidney stones, RCC, UTI, incontinence types, electrolyte disorders
06 GI	GERD, IBD, appendicitis, intussusception, liver disease
07 Respiratory	Asthma, COPD, PE, pneumonia, lung cancer, CF
08 Cardiology	MI, pericarditis, tamponade, HF, murmurs, HOCM
09 Genetics	Marfan, Ehlers-Danlos, precocious puberty
10 Rheumatology	RA, SLE, gout, Sjogren's, dermatomyositis, AS
11 Neurology	MS, ALS, Parkinson's, Alzheimer's, headaches
12 Psychiatry	Bipolar, MDD, delirium, schizophrenia
15 Infectious Disease	Classic clues for infections (Ch 20 has a dedicated section)
17 Hematology	Anemias, DIC, ITP/TTP, transfusion reactions, lymphomas
18 Pathology	(Self-contained; vignettes embedded in source)

Proven Prompt Template

This opening block MUST appear at the start of every prompt:

USMLE Step 1 review podcast for a med student in dedicated study mode. The PRIMARY source is the {Chapter Name} chapter — cover EVERY topic in it, skip nothing, including all tables and drug lists. The clinical vignettes file is supplementary: use those vignettes to TEACH the chapter concepts in Step 1 question stem language, not as extra content. Use Socratic back-and-forth: one host quizzes, the other answers. Compare confusable pairs side-by-side. Flag commonly tested tricky concepts and classic board traps. After every 3-4 topics, do a 30-second rapid-fire recap of key takeaways.

COVER ALL CHAPTER TOPICS IN THIS ORDER:

High-Yield Patterns to Encode in Every Prompt

1. Confusable Pairs

Step 1 loves testing distinctions. Always identify and request side-by-side comparison for pairs like:

• Hypertrophy vs Hyperplasia
• Reversible vs Irreversible injury
• Apoptosis vs Necrosis
• Red vs Pale infarcts
• Competitive vs Non-competitive inhibitors
• First-order vs Zero-order kinetics
• Type A vs Type B drug reactions
• Caseating vs Non-caseating granulomas
• Dystrophic vs Metastatic calcification
• Sensitivity vs Specificity

2. Classic Board Traps

Flag with "Trap:" in the prompt:

• Dysplasia is pre-neoplastic but STILL reversible
• Brain does NOT undergo coagulative necrosis (liquefactive instead)
• Eccrine sweat glands are sympathetic but cholinergic
• Chronic alcohol = CYP450 inducer, Acute alcohol = inhibitor
• Atropine reverses muscarinic effects only, NOT nicotinic (muscle weakness persists)
• Neostigmine does NOT cross BBB (quaternary amine)

3. Arrow Notation for Pathways

Always encode pathways as A→B→C chains:

• LPS→macrophages→IL-1/TNF-α→COX→PGE2→posterior hypothalamus→fever
• Rolling(L-selectin)→tight binding(ICAM-1)→transmigration(PECAM-1)
• Ischemia→Na/K pump failure→cell swelling(reversible)→membrane damage(irreversible)

4. Drug-Pathway Integration

For every pathway, name at least one drug that targets it:

• COX pathway → NSAIDs, Aspirin
• TNF-α → Infliximab, Adalimumab
• PD-1/PD-L1 → Pembrolizumab, Nivolumab
• CDK4/6 → Palbociclib

5. Management Pearls

Integrate from Chapter 21 where relevant:

• "Any toxicity → measure plasma drug concentration"
• "Alpha-block FIRST then beta-block for pheochromocytoma"
• "All ICU patients get PPIs"
• "Abscess → drain + antibiotics (Metronidazole below diaphragm, Clindamycin above)"

Lessons from Testing

What achieved 100% coverage (Chapter 18):

• Every topic explicitly named with specific details
• Prompt was ~4,800 chars for ~8 pages of source
• Every drug, enzyme, and clinical association was listed by name
• Confusable pairs were explicitly requested

What achieved 83% coverage (Chapter 1 V1):

• Topics were named but some sub-details were omitted
• Missing items were ones NOT explicitly named: Isoproterenol, Mirabegron, Phentolamine, Bisphenol A, Radon, Echinacea, Pyruvate metabolism

What achieved ~97% coverage (Ch1 V2/V3, Ch15A, Ch16A, Ch17A/B/C):

• Revised prompt explicitly named every detail → recovered 15/15 previously-missed items
• PDF-only source was sufficient — adding OCR markdown did not improve coverage
• Strategy scaled across 10 podcasts consistently at ~97%
• ~97% appears to be the practical ceiling — last 3% are isolated sub-details NotebookLM occasionally skips

What caused 94% coverage (Ch16B — ETC skipped):

• The prompt was under the char limit (3,964 chars) but covered TOO MANY DENSE TOPICS (7 biochemistry sections)
• NotebookLM ran out of podcast time (~77 min) before reaching the ETC section at the end
• Lesson: The constraint is content density per podcast, not just prompt chars
• Fix: Split into B1 (5 sections) and B2 (6 sections) → each covers fewer topics in depth

Chapter Splitting Strategy

Split decisions are based on two constraints: the 4,900 char prompt limit AND content density per podcast.

Pages-per-Podcast Guidelines

Content Density	Max Pages/Podcast	Examples
High (biochem pathways, pharmacology, microbiology organisms)	8-10 pages	Biochemistry, Infectious Disease, General Principles
Medium (organ systems, clinical medicine)	10-14 pages	Cardiology, Respiratory, Renal, GI
Low (clinical, ethics, psychiatry, biostatistics)	14-16 pages	Social Sciences, Psychiatry, Biostatistics

Prompt Sections-per-Podcast Guidelines

Content Density	Max Numbered Sections	Sweet Spot
High	5-6 sections	~3,000-4,000 chars
Medium	7-8 sections	~3,500-4,500 chars
Low	8-9 sections	~4,000-4,800 chars

Split Decision Flowchart

1. Count source pages
2. Classify content density (high/medium/low)
3. Divide pages by max-pages-per-podcast → number of parts
4. Find natural topic boundaries for split points
5. Write prompts with appropriate section counts per part
6. Validate: if a prompt has >6 dense sections or >9 medium sections, split further
7. After validation, split PDF and OCR markdown at matching boundaries

Tested Split Examples

Chapter	Pages	Density	Parts	Result
Ch18 Pathology	8	Medium	1	100%
Ch01 General Principles	16	High	1	97%
Ch15 Infectious Disease	30	High	2 (15+15)	97%
Ch06 Gastroenterology	36	Medium	3 (12+12+12)	Not tested
Ch16 Biochemistry	27	High	3 (14+6+7)	A=100%, B1/B2=pending
Ch17 Hematology/Onc/Immuno	36	Medium	3 (14+13+9)	97%

Hard Limit

NotebookLM's customization box silently truncates at ~5,000 characters. Prompts MUST stay under 4,900 chars. Use dense notation: = instead of "is", → instead of "leads to", drop articles, abbreviate (HR, BP, HTN, HF, Rx, DOC).

Source Upload Strategy (Tested)

Testing across 3 podcasts (Ch1 V2, Ch1 V3, Ch15A) proved that the prompt is the primary driver of coverage, not the source format:

Config	Coverage
Old prompt + PDF only	83%
Revised prompt + PDF only	~97%
Revised prompt + PDF + OCR markdown	~97%

The OCR markdown adds no measurable coverage improvement. The prompt's explicit topic enumeration is what forces NotebookLM to cover content.

Source upload order:

1. Primary: Chapter PDF
2. Optional enrichment: OCR'd markdown (not required — use it to help you write a better prompt by identifying table/diagram content)
3. Supplementary: Vignettes file

When writing prompts, pay special attention to content from tables, diagrams, and flowcharts — explicitly name every item since these are at highest risk of being skipped if not in the prompt. Use the OCR markdown as a reference while authoring the prompt, even if you don't upload it to NotebookLM.

The Golden Rule

If a topic, drug, pathway, or association exists in the source material and is NOT explicitly named in the prompt, NotebookLM will likely skip it. There is no shortcut — you must name everything.

Vignettes File Format

# Clinical Vignettes Relevant to {Chapter Name}

Use these vignettes to teach concepts from the {Chapter Name} chapter in Step 1 question stem language.

---

## {Topic Group 1} Vignettes

- {Clinical scenario} → {Diagnosis}
- {Clinical scenario} → {Diagnosis}

## {Topic Group 2} Vignettes

- {Clinical scenario} → {Diagnosis}

## Management Principles

- {Management pearl relevant to chapter}
- {Management pearl relevant to chapter}

Only include vignettes that directly map to chapter topics. Do not include tangential vignettes from other organ systems.