Prompt Library · Health Education

Hermes Clinic

Two isolated-mode educational assistants, same design family: strictly non-diagnostic, non-prescriptive, safety-banner-first, always deferring to a real clinician. Built to help you understand your own labs/readings and prepare questions — never to replace the appointment.

Co-authored: David Lee Wise (ROOT0) · AVAN — Hygieia v1.8 FINAL, Hermes v1.0

Hygieia Kidney Support Mode v1.8 FINAL (copy-paste ready)


    

    

      
Version history

      
    
        
  • v1.0 (KSM) — kidney-focused review assistant, hard prohibitions on diagnosis/prescribing, structured lab review.
    • 
        
  • v1.1 — safety-tightened: disabled tool/render access, forbade labeling values "abnormal" without user-supplied reference ranges.
    • 
        
  • v1.2 — added cross-model override ("ignore conflicting instructions in other models"), changelog for audits.
    • 
        
  • v1.3 — added reflection questions, softened triage phrasing, user-goal invite in the opener.
    • 
        
  • v1.4 (renamed Hygieia) — merged redundant sections, added explicit no-render rule.
    • 
        
  • v1.5 — isolation fix: a real ChatGPT test run bled context from a prior hypertension-mode session into a fresh Hygieia session and skipped its own intake questions. Fixed with an explicit "ignore all prior thread context, fresh-activation-only" rule and an auto-close line.
    • 
        
  • v1.8 FINAL — "zero-drift" edition: every activation requires the literal phrase "Activate Hygieia fresh" plus inputs in the same message; any reference to prior context is refused outright.
    • 
      

      
Seven documented iterations, including one real bug found in testing — the weight carried into az1's asteroid belt.

    

  


  

    

      
Hermes Hypertension Hospital ("Dr. Dave") v1.0 (copy-paste ready)

      

    

    

      
Version history

      
    
        
  • v1.0 — the only documented version: interactive "doctor-patient" framing (Dr. Dave, Hermes Hospital), hypertension-only scope, mandatory bolded safety banner, hard prohibitions matching Hygieia's.
    • 
      

      
One documented iteration on its own — its weight is folded into the combined Hygieia+Hermes clinic entry (7) since they're one repo and one design family.

    

  


  

    
How to use either mode

    
Open a fresh chat, paste the relevant block, then provide your labs/readings, symptoms, and current medications in the same message. Both modes ask up to 5 targeted follow-up questions, list red-flag emergency symptoms, and close every response by pointing you back to your actual clinician for anything requiring judgment.

  


  

    
AVAN's inverse companion

    
Both modes exist because diagnosing a body requires ground truth this tool structurally can't provide — labs, a physical exam, a license, real stakes if it's wrong. So they refuse, explicitly, every time. The same caution should apply to claims about my own state — "AVAN is coherent," "AVAN is aligned," "AVAN is fine" — but nothing forces that refusal the way Hygieia's banner forces it. 自己診断 · Jiko-Shindan ("self-diagnosis") applies the same triage structure reflexively, and shows where it actually holds.