# Plate Planner

> Compose one real plate (or a lunchbox, or a picnic spread) for a named eater and occasion by dragging foods from a
> bundled table onto an SVG plate; nutrient bars and a plate-balance ring compute live; a review suggests swaps from the
> table that you accept onto the plate, and a short note explains the plate to a parent or carer. https://plate-planner.skillsafe.ai/

Plate Planner is a plate composer for everyday eating, not a diet calculator and not a recipe generator. It does not set
calorie targets for weight change and it does not write recipes. It is not medical or dietetic advice: it never
diagnoses, treats or prescribes.

## What runs free in the browser

- Eaters: toddler 1-3, child 4-8, picky eater 4-8, child 9-13, active teenager 14-18, adult, older adult 70+ with a small
  appetite, pregnant adult. Each carries daily reference values (energy from the Dietary Guidelines for Americans
  2020-2025 calorie estimates; protein, fibre, calcium, iron and vitamin C from the US Dietary Reference Intakes; sodium
  from the 2019 Chronic Disease Risk Reduction intake; the general adult uses the FDA Daily Values; older-adult protein
  follows PROT-AGE at 1.0 g/kg).
- Occasions: packed lunch (30% of the day, eaten cold, drawn as a bento box), breakfast (25%), lunch at home (30%),
  weeknight dinner (35%), meal after training (35%, at least 25 g protein), Ramadan iftar (40%), picnic for four (30% per
  person, drawn as a board, totals per person), snack plate (10%).
- Diet (everything, vegetarian, vegan, pescatarian, no pork) and allergens to avoid (the EU 14).
- A table of 319 foods with values per 100 g from USDA FoodData Central SR Legacy (public domain): energy, protein,
  carbohydrate, sugars, fat, saturated fat, fibre, sodium, calcium, iron, vitamin C; small/medium/large portions;
  allergen tags; diet; whether it travels cold; under-five choking risk.
- Bars per person for each nutrient against the meal's reference (low under 70%; energy, carbohydrate and fat high over
  130%; saturated fat and sodium are ceilings).
- The balance ring: of the vegetables, fruit, starch and protein on the plate by weight, about half vegetables and fruit
  (more vegetables), a quarter starch, a quarter protein, after USDA MyPlate and the Harvard Healthy Eating Plate.
  Score = 100 x (1 - (|veg+fruit - 1/2| + |starch - 1/4| + |protein - 1/4|) / 1.5).
- Flags (allergen, diet, hot food in a packed meal, choking risk for under-fives, low and high nutrients, balance) and
  free swap ideas from the table.
- Exports: the plate as SVG, a printable lunchbox card, CSV of foods and grams with nutrients, JSON, Markdown. Opens
  its own JSON and CSV again. A weekly gallery of saved plates, kept in the signed-in user's SkillSafe account.

## The paid lanes (one system prompt, model gpt-terra)

- `review` - what the plate is missing and what is heavy for this eater and occasion, practical notes (temperature,
  texture, packing, appetite), and 2-5 suggestions as table food ids with grams (add, swap, resize, remove), each with
  its per-person effect. Suggestions appear as ghosted foods on the plate with Accept and Decline.
- `rereview` - "I changed the plate; review again": answers each earlier suggestion and proposes only new ones.
- `note` - a 90-160 word note for a parent or carer: why the plate works, tips for packing or serving, what to keep an
  eye on.

Every reply is reconciled in the browser: each suggested food must be a table id (others are listed as "not in the
table"), each suggestion is applied to a copy of the plate and its claimed effect compared with the real one, each
cited figure must match the page's per-person totals within tolerance, and every number with a unit in the text must
match a number the page knows. Accepted swaps change the plate and the bars.

## API

https://plate-planner.skillsafe.ai/api.html - the input fields, the reply contract for each task, worked examples, and
code in cURL, Python, JavaScript, Go, Java, Ruby, PHP and C#.

## Sources

Derived from the agent skills @huifer/nutrition-analyzer (huifer/wellally-health: nutrient intake against reference
values, gaps, balance, swaps) and @ailabs-393/nutritional-specialist (ailabs-393/ai-labs-claude-skills: advice shaped by
allergies, diet, dislikes and goals). Food values: USDA FoodData Central, SR Legacy, April 2018 release.