This is a submission for the Runner H "AI Agent Prompting" Challenge
🧠 Concept Overview
A multi-agent AI system that:
- Plans meals based on dietary preferences, goals, and schedule.
- Flags harmful food combinations (e.g., dairy + citrus, iron + calcium blockers).
- Warns against foods based on medical conditions (e.g., “avoid grapefruit if on statins” or “no nightshades for arthritis”).
✅ Why It’s Unique & Useful
- Combines meal planning, dietary safety, and personalized health guidance.
- Solves a real-world daily problem — what to eat safely, deliciously, and conveniently.
- Useful for people with diabetes, hypertension, thyroid issues, heart-issues, PCOS, alergic, or on specific meds.
- Stands out by not only recommending what to eat, but also what not to combine or consume.
What I Built
NutriGuard AI is a prompt-engineered multi-agent workflow designed to provide personalized meal planning, toxic food combination detection, and health-condition-aware dietary filtering.
This Runner-H AI system solves a common yet critical problem:
“What should I eat that’s safe, healthy, and suited to my condition?”
NutriGuard allows users to input dietary preferences (e.g., vegetarian, keto), medical conditions (e.g., heart-issues, thyroid, PCOS, diabetes), and dislikes/allergies. The system:
- Creates a smart daily or weekly meal plan
- Warns about harmful food pairings
- Flags ingredients that may worsen the user's health condition
- Generates a shopping list with safe substitutions
Each function is powered by dedicated prompt-driven agents, making the system modular, transparent, and adaptable.
Demo
Example User Flow
User Input:
“I’m vegetarian with thyroid issues. I don’t like mushrooms. Plan my meals for the next 3 days, and avoid foods that interfere with my medication.”
Output:
- Meal Plan:
- ✅ Breakfast: Oatmeal with almonds and berries (no soy or broccoli)
- ⚠️ Lunch: Lentil salad with lemon — replaced spinach (high in oxalates)
- ✅ Dinner: Paneer tikka with quinoa and turmeric vegetables1. Actual numbers don't matter, just that it's a number
- 🛑 Warning: Avoid soy, cabbage, and cauliflower — they can suppress thyroid function.
- 🛒 Shopping List: Includes safe alternatives and quantities.
Screenshot
Link
https://runner.hcompany.ai/chat/4819c15f-43ca-4f18-83a9-63855ba4539d/share
How I Used Runner H
I used Runner H to build a multi-step, prompt-based automation where each stage of the meal planning process is handled by a dedicated AI agent:
Agents:
- 🧠 Meal Planner Agent – Generates meals based on user preferences, schedule, and dietary needs.
- ⚠️ Toxic Combo Checker – Detects food pairings that may be harmful (e.g., milk + citrus).
- 🚫 Condition Filter Agent – Flags and substitutes foods harmful to medical conditions.
- 🛒 Shopping List Agent – Compiles a clean, ingredient list optimized for nutrition and safety.
- 📚 Nutrition Explainer (Optional) – Explains flagged items and nutritional choices.
Workflow in Runner H:
- Each step is chained using clear prompt-based tasks.
- User input is parsed once, then passed through each agent.
- Agents respond with structured JSON blocks for easy integration with UI or PDF generation.
Sample Prompt for the Toxic Combo Agent
You are a food safety expert. Analyze the above-provided meal and flag
any known harmful food combinations. Use evidence-based knowledge
(e.g., high-iron foods mixed with calcium inhibit absorption,
citrus + milk may cause digestion issues, etc.).
Sample Prompt for Condition Filter Agent
You are a clinical nutritionist. Based on the condition: “Hypothyroidism”,
scan the above-provided ingredients and remove or replace any foods
that negatively affect this condition or common medications related
to it. Explain your choices briefly.
This modular approach allows other users to replicate, extend, or plug in their own use cases (e.g., low-FODMAP diet, PCOS-safe foods).
Use Case & Impact
NutriGuard AI is built for:
- Individuals with chronic conditions (diabetes, thyroid issues, hypertension, PCOS)
- Health-conscious eaters seeking personalized meal planning
- Caregivers or nutritionists creating plans for others
- Anyone overwhelmed by conflicting online food advice
Real-World Impact:
- Reduces the risk of harmful food interactions
- Ensures condition-specific dietary compliance
- Makes meal planning easy and safe
- Saves time and avoids misinformation
It’s like having a personal dietitian + food scientist + chef all in one AI flow.
Social Love
Sharing this on X/Twitter and LinkedIn
Top comments (0)