Practice Questions

1e. Image library (optional) Read public/media/{Notebook}/images/ATTRIBUTIONS.json. Filter for images whose pages array includes the current slug. Available for image-based questions.

Step 2: Identify Coverage Gaps

• Map existing questions to content sections using tags and stems
• Check style distribution (target: 30% concept-check, 25% mechanism, 25% compare-contrast, 10% finding-based, 10% vignette)
• Check domain distribution
• Cross-check learning objectives — any untested?
• Identify lecture-exclusive content not fully covered

Present the gap analysis before generating. Example: "12 existing questions. Sections X and Y uncovered. 'mechanism' style has only 1 question. Objectives 3 and 7 untested. Suggesting 8 new questions."

Step 3: Generate Questions

Follow these principles:

• Test understanding, not recall — scenarios, not definitions
• Every wrong answer teaches — rationales explain real concepts
• Clinical vignettes are realistic — specific ages, lab values, relevant negatives
• Distractors are plausible — close-but-wrong options testing nuance
• Distribute correctIndex evenly across 0–4. Hard rule: no single index may exceed 40% of questions. QuizHub shuffles at runtime, but even distribution is required for data quality

See quiz schema for the exact JSON format.

Image-based questions: When relevant images exist, add "image" field. Preferred format: { "src": "/media/...", "alt": "...", "caption": "..." }. QuizHub also accepts a plain string path. Verify the image file exists in public/ before referencing. Reference in stem: "A peripheral blood smear is shown [see image]..." Use finding-based style for most.

Step 4: Validate & Merge

1. Schema compliance (all required fields, valid style/domain)
2. ID uniqueness (continue from highest existing)
3. correctIndex distribution
4. correctIndex ↔ rationale alignment: For EVERY question, verify rationale[correctIndex] starts with "CORRECT" and all others start with "INCORRECT". This is the #1 source of quiz bugs.
5. Image path verification: If image field exists, verify the file exists in public/
6. Factual accuracy against glossary
7. Append to existing array — do NOT modify or reorder existing questions
8. Run post-edit validation script (see below)

Write the complete updated quiz file and present a summary: questions added, gaps filled, updated distribution.

Post-Edit Validation (REQUIRED)

After writing the quiz file, run:

node -e "
const fs=require('fs'),p=require('path');
const dir=p.join(process.cwd(),'src/data/quizzes');
let bad=0;
(function walk(d){fs.readdirSync(d).forEach(f=>{const fp=p.join(d,f);if(fs.statSync(fp).isDirectory())return walk(fp);if(!f.endsWith('.json'))return;const qs=JSON.parse(fs.readFileSync(fp,'utf8'));(Array.isArray(qs)?qs:qs.questions||[]).forEach(q=>{if(!q.rationale)return;for(const[k,v]of Object.entries(q.rationale)){if(k===String(q.correctIndex)&&!/^Correct/i.test(v)){console.log('MISMATCH',q.id,'correctIndex='+q.correctIndex,'but rationale['+k+'] says:',v.slice(0,60));bad++;}if(k!==String(q.correctIndex)&&/^Correct/i.test(v)){console.log('MISMATCH',q.id,'rationale['+k+'] says Correct but correctIndex='+q.correctIndex);bad++;}}if(q.image){const src=typeof q.image==='string'?q.image:q.image.src;if(src&&!fs.existsSync(p.join(process.cwd(),'public',src.replace(/^\//, ''))))console.log('MISSING IMAGE',q.id,src);}});})})(dir);
console.log(bad?bad+' issues found':'All clear');
// Distribution check
(function checkDist(d){fs.readdirSync(d).forEach(f=>{const fp=p.join(d,f);if(fs.statSync(fp).isDirectory())return checkDist(fp);if(!f.endsWith('.json'))return;const qs=JSON.parse(fs.readFileSync(fp,'utf8'));const arr=Array.isArray(qs)?qs:qs.questions||[];if(arr.length<5)return;const counts={};arr.forEach(q=>{const ci=q.correctIndex;counts[ci]=(counts[ci]||0)+1;});for(const[idx,cnt]of Object.entries(counts)){if(cnt/arr.length>0.4)console.log('BIAS',f,arr.length+'q',Math.round(cnt/arr.length*100)+'% at index '+idx);}});})(dir);
"

If issues are found, fix them before finishing.

Edge Cases

• No existing quiz file: Create new file, start IDs at 1
• User requests specific styles only: Skip full gap analysis, generate requested styles
• Large page (>500 lines JSX): Focus on section headers, clinical pearls, comparison tables, warning boxes
• Glossary vs. page discrepancy: Flag to user, do not silently resolve