Fuel Check

# Fuel Check

Context-aware food analysis. Takes a photo, identifies what you're eating, and scores it against your body's current state — steps, heart rate, weather, time of day.

## Arguments

`<image_path>` — Absolute path to a food photo. If no argument, use the most recent image in `~/.claude/channels/telegram/inbox/` (sorted by filename timestamp).

Examples:
- `/fuel-check` — analyze the last photo sent via Telegram
- `/fuel-check /path/to/photo.jpg` — analyze a specific image

## Workflow

### Step 1: Resolve Image

- If argument provided, use it as the image path
- If no argument, find the most recent `.jpg` or `.png` file in `~/.claude/channels/telegram/inbox/` by sorting filenames descending (filenames are timestamp-prefixed)
- If no images found, tell the user to send a food photo via Telegram first

### Step 2: Read and Verify

- Read the image file
- Determine if the image contains food or drink
- If NOT food/drink: reply "That's not food — send me something you're eating or drinking." and stop
- If it IS food/drink: identify what it is (be specific — brand, flavor, size if visible)

### Step 3: Gather Context (in parallel)

Call all three in parallel:
- `mcp__garmin-connect__get_daily_stats` with today's date
- `mcp__garmin-connect__get_heart_rate` with today's date
- `mcp__open-meteo__weather_forecast` or `mcp__open-meteo__weather_archive` for the user's location (Iowa City: lat 41.66, lon -91.53) for today

Also note the current time of day.

### Step 4: Identify and Estimate

State the food/drink item clearly, then provide:
- Approximate calories
- Key macros (protein, carbs, fat, sugar, sodium — whatever is relevant)
- Serving size estimate from the photo

### Step 5: Score (6 Dimensions)

Present a table scoring each dimension out of 10, with a one-line justification that references the context:

| Category | Score | Why |
|---|---|---|
| **Context Requirement** | X/10 | Does your body need this right now based on today's activity, weather, and time? |
| **Health** | X/10 | Nutritional quality — vitamins, whole ingredients vs processed, sugar load |
| **Satiety** | X/10 | Will this actually make you feel full? Protein and fiber help, liquid calories don't |
| **Taste** | X/10 | How good does this probably taste, especially given your current state (depleted, rested, etc)? |
| **Bloat** | X/10 | Higher = less bloat. Will this sit well, especially if you're about to walk? |
| **Addiction** | X/10 | How engineered is this to make you want more? Sugar + salt combos, hyperpalatable design |

Scoring guidelines:
- Context Requirement: factor in steps so far today, Body Battery, weather (heat = electrolyte need), time of day (late night sugar is worse), whether they've eaten today
- Health: whole foods score high, processed/sugary score low. Electrolyte drinks get partial credit when genuinely needed
- Satiety: solid food with protein/fiber scores high, liquid calories score low
- Taste: be honest. Junk food tastes good. Depleted bodies make sweet/salty taste even better
- Bloat: liquid and simple foods score high (less bloat). Heavy meals, dairy, carbonation score lower. Context matters — post-dinner walking is worse
- Addiction: sugar + salt + fat combos score high (more addictive). Whole foods score low. Engineered products (fast food, sports drinks, candy) score higher

### Step 6: The Real Read

End with 2-3 sentences of opinionated, contextual analysis. This is NOT generic nutrition advice. Reference specific data points — today's steps, Body Battery, the weather, the time, what they've eaten (or haven't eaten) today. Be direct. No hedging.

Match tone to the situation:
- If they're eating junk after a big effort day: acknowledge the craving but call out the pattern
- If they're eating well and it matches their needs: validate it
- If they're eating at a weird hour: flag it
- If they skipped meals earlier: connect the dots between that and what they're reaching for now