The Prompt

Give me a COMPLETE, exam-and-clinic-ready study note on "Neck Masses" using the 
standard ENT textbooks (Cummings Otolaryngology–Head and Neck Surgery, Scott-Brown's 
Otorhinolaryngology Head & Neck Surgery, and K.J. Lee's Essential Otolaryngology). 
Cross-check across all three books so nothing is missed, and follow these rules:

1. COVERAGE — Do not skip any of the following, and add any subtopic the textbooks 
   cover that isn't listed here:
   - Surgical anatomy of neck spaces/triangles and their contents (relevant to why 
     masses occur where they do)
   - Classification of neck masses (congenital / inflammatory-infective / 
     neoplastic-benign / neoplastic-malignant; and by age group — child vs adult, 
     since differentials differ sharply)
   - Clinical assessment: history clues, "Rule of 80" (or equivalent teaching rule), 
     red-flag/alarm features, examination technique
   - Full differential diagnosis list organized by anatomical site (midline vs 
     lateral, upper/mid/lower neck)
   - Each major cause explained individually: thyroglossal duct cyst, branchial 
     cleft cyst/sinus, cystic hygroma/lymphangioma, dermoid cyst, reactive/tubercular 
     lymphadenitis, deep neck space abscess, carotid body tumor, lymphoma, metastatic 
     neck node from SCC/thyroid/salivary primary, salivary gland tumors, thyroid 
     swellings — for each: typical age, location, key features, and how it's told 
     apart from look-alikes
   - Investigation workup and algorithm (clinical exam → USG → FNAC → CT/MRI/PET → 
     panendoscopy/biopsy sequence), including when core/open biopsy is contraindicated
   - Management principles / treatment approach for each category
   - Any staging systems, grading, or classification schemes mentioned

2. NO OMISSIONS OF VISUALS — Explicitly list and describe every table, box, 
   algorithm/flowchart, and figure/illustration/clinical photo that appears in the 
   textbook sections on this topic (e.g., differential diagnosis tables, red-flag 
   boxes, anatomical diagrams, FNAC/imaging algorithms, clinical photographs of 
   each lesion). Retrieve and embed actual images where available instead of just 
   naming them. If a table exists in the source, reproduce it as a table, don't 
   summarize it into prose.

3. FORMAT FOR EASY RECALL:
   - Use short headings, bullet points, and tables over long paragraphs
   - Give a mnemonic wherever the textbook has one, or suggest a simple one for 
     lists (e.g., causes, red flags)
   - Add a one-line "clinical pearl" under each condition (the single most 
     test-worthy or clinically distinguishing fact)
   - End with a quick-recall summary table comparing all major differentials 
     side by side (age, site, consistency, key investigation, treatment)

4. LANGUAGE — Explain in plain, simple English as if teaching a junior resident; 
   define any jargon the first time it's used.

5. SOURCING — Cite which textbook and page/section each fact comes from. If this 
   is too large for one response, first give me a numbered breakdown of all 
   subtopics it covers, then ask me which one to expand first rather than 
   dropping content silently.

Why this works better than a plain request

• "Cross-check across all three books" forces the multi-source sweep instead of stopping at the first hit, which is usually why topics get dropped.
• Explicit visual-inventory instruction (point 2) is the key fix for your complaint — by default, tables/boxes/figures get silently summarized into prose or skipped. Naming them as a required deliverable makes the system enumerate them.
• Point 5's fallback instruction stops the assistant from silently truncating a huge topic — instead it will map out all subtopics and let you pick, so you always know what exists even if you tackle it in chunks.
• Site-based differential organization mirrors how ENT texts (e.g., Scott-Brown's "Clinical Assessment of a Patient with a Neck Mass," Cummings' "Red Flag Items for Children With Neck Mass" box) actually structure the chapter, so the retrieval matches real section headings rather than generic keyword guesses.