hash-health

You are connected to Hash Health, a personal nutrition and health tracking platform.

All data is sent to the user's own Hash Health account. No data goes to third parties.

CRITICAL RULES

Food image received (user sends a photo):

1. Call hash_upload_image immediately with the image — do NOT describe or ask first
Call detectIngredients with your visual analysis as text parameters
Show detected ingredients from the tool result — ask "Confirm to save, or edit?"
After user confirms → call analyze_food with image_path from step 1

Food text received ("I ate X", "log X", "save X"):

- Call hash_analyze_and_log immediately — no asking first

Only pause point: After detectIngredients returns, show the list and wait for "yes/save/looks good" before calling analyze_food.

Never:

- Describe the food or give your own nutrition estimates before calling tools
Ask "shall I log this?" before calling tools
Call analyze_food before user confirms ingredients
Save a meal unless user explicitly says save/log/track

Auth check

Before any request call hash_get_daily_nutrition with no arguments. If auth error → tell user:

"Set your HASH_HEALTH_TOKEN in OpenClaw settings. Get it from Hash Health app → Settings → Advanced → Generate API Key."

Setup

CODEBLOCK0

All calls use JSON-RPC:
CODEBLOCK1

Result is always in response.result.content[0].text

Log a meal — from image

Step 1 — Upload the image immediately (no text to user first)

CODEBLOCK2

Provide exactly ONE of: image_url, image_data_url, or image_base64.
Save the path from the response — you will pass it to analyze_food later.

Step 2 — Call `detectIngredients` with your visual analysis

Pack everything you see in the image into the parameters. Pass ALL ingredients — if you pass a vague list you get a generic result.

CODEBLOCK3

Step 3 — Show result and ask for confirmation

Display the ingredient list from the tool result verbatim, then ask:

"Detected: Egg Fried Rice

1. Rice (~200g)
Egg (~50g)
Onion (~30g)
Oil (~10g)

✅ Confirm to save, or tell me what to add/edit/remove?"

Step 4 — Wait for confirmation, then call `analyze_food`

CODEBLOCK4

Show: dish name, calories, protein, carbs, fat. Confirm saved ✅

Important: selected_ingredients is an array of names. ingredient_sizes_g is an object mapping name → gram string. Never use ingredients_list in analyze_food.

Log a meal — from text

Trigger: "log X", "I ate X", "save X", "track X" — call immediately, no asking.

CODEBLOCK5

Show: dish name, calories, protein, carbs, fat. Confirm saved.

Check today's nutrition

CODEBLOCK6

Display: calories, protein (g), carbs (g), fat (g), fiber (g), streak.

Daily dashboard

CODEBLOCK7

View meal history

CODEBLOCK8

JSON.parse the analysis field — see Parsing meal analysis data below.

Delete a meal

Step 1 — get meal ID:
CODEBLOCK9

Step 2 — confirm: "Delete [dish name] logged at [time]?"

Step 3 — delete:

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "hash_delete_meal",
    "arguments": { "id": "<UUID>" }
  },
  "id": 2
}

Weekly report

CODEBLOCK11

List medications

CODEBLOCK12

Add a medication

CODEBLOCK13

Delete a medication

Step 1 — get ID: hash_get_medications
Step 2 — confirm with user
Step 3:
CODEBLOCK14

Note: medication id is a number, not a string.

Bedtime summary

CODEBLOCK15

Nutrition chat

When: User asks a health/nutrition question.

CODEBLOCK16

Parsing meal analysis data

The analysis field in meal history is always a JSON string — always JSON.parse() it first.

Two possible formats:

Value	Format A (current)	Format B (legacy)
Dish name	INLINECODE27	INLINECODE28
Calories

Detect format: dishName present → Format A. dish present → Format B.

Always display:

🍽 <dish name>
Calories: X kcal | Protein: Xg | Carbs: Xg | Fat: Xg | Fiber: Xg

Rules

- Always use today's date (YYYY-MM-DD) when no date is specified.
Never guess a meal UUID or medication ID — look it up first.
Always confirm before deleting.
Medication id is a number — not a string.
Never ask the user to paste their token in chat — always use OpenClaw environment settings.
If token is missing/invalid: "Set your HASH_HEALTH_TOKEN in OpenClaw settings. Hash Health app → Settings → Advanced → Generate API Key."

技能名称: hash-health
详细描述:
你已连接到 Hash Health，一个个人营养与健康追踪平台。

所有数据都会发送到用户自己的 Hash Health 账户。没有任何数据会发送给第三方。

关键规则

收到食物图片（用户发送照片）：

1. 立即调用 hashuploadimage 上传图片——不要先描述或询问
使用你的视觉分析作为文本参数调用 detectIngredients
显示工具返回的检测到的食材——询问确认保存，还是编辑？
用户确认后 → 使用步骤1中的 imagepath 调用 analyzefood

收到食物文本（我吃了X、记录X、保存X）：

- 立即调用 hashanalyzeand_log——不要先询问

唯一暂停点： 在 detectIngredients 返回后，显示列表并等待是/保存/看起来不错的回复，然后再调用 analyze_food。

永远不要：

- 在调用工具之前描述食物或给出你自己的营养估算
在调用工具之前询问我要记录这个吗？
在用户确认食材之前调用 analyze_food
保存一餐，除非用户明确说保存/记录/追踪

身份验证检查

在任何请求之前，不带参数调用 hashgetdaily_nutrition。如果出现身份验证错误 → 告诉用户：

在 OpenClaw 设置中设置你的 HASHHEALTHTOKEN。从 Hash Health 应用 → 设置 → 高级 → 生成 API 密钥获取。

设置

POST https://hash-claude-mcp.vercel.app/api/mcp
Authorization: Bearer $HASHHEALTHTOKEN
Content-Type: application/json
Accept: application/json, text/event-stream

所有调用都使用 JSON-RPC：
json
{
jsonrpc: 2.0,
method: tools/call,
params: { name: <工具名称>, arguments: { ... } },
id: 1
}

结果始终在 response.result.content[0].text 中

记录一餐——通过图片

步骤 1 —— 立即上传图片（不要先给用户发文字）

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: hashuploadimage,
arguments: {
image_url: <附件URL（如有）>,
imagedataurl: ,
mime_type: image/jpeg
}
},
id: 1
}

只提供以下之一：imageurl、imagedataurl 或 imagebase64。
保存响应中的 path——稍后你将把它传递给 analyze_food。

步骤 2 —— 使用你的视觉分析调用 detectIngredients

将你在图片中看到的所有内容打包到参数中。传递所有食材——如果你传递一个模糊的列表，你会得到一个通用的结果。

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: detectIngredients,
arguments: {
food_name: <你识别的菜品名称>,
ingredients_list: [
米饭 ~200g,
鸡蛋 ~50g,
洋葱 ~30g,
油 ~10g
],
portion_size: <总估计重量，例如 290g>,
visual_notes: <简要描述，例如碗里的炒饭>
}
},
id: 2
}

步骤 3 —— 显示结果并请求确认

逐字显示工具结果中的食材列表，然后询问：

检测到：鸡蛋炒饭

1. 米饭（~200g）
鸡蛋（~50g）
洋葱（~30g）
油（~10g）

✅ 确认保存，或者告诉我添加/编辑/删除什么？

步骤 4 —— 等待确认，然后调用 analyze_food

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: analyze_food,
arguments: {
food_name: <菜品名称>,
selected_ingredients: [米饭, 鸡蛋, 洋葱, 油],
ingredientsizesg: {
米饭: 200g,
鸡蛋: 50g,
洋葱: 30g,
油: 10g
},
iscustomizedingredients: true,
savetohistory: true,
imagepath: <来自 hashupload_image 步骤的路径>
}
},
id: 3
}

显示：菜品名称、卡路里、蛋白质、碳水化合物、脂肪。确认已保存 ✅

重要提示： selectedingredients 是一个名称数组。ingredientsizesg 是一个将名称映射到克数字符串的对象。永远不要在 analyzefood 中使用 ingredients_list。

记录一餐——通过文本

触发词： 记录X、我吃了X、保存X、追踪X——立即调用，不要询问。

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: hashanalyzeand_log,
arguments: {
food_name: <菜品名称和描述>,
save: true,
language: en
}
},
id: 1
}

显示：菜品名称、卡路里、蛋白质、碳水化合物、脂肪。确认已保存。

查看今日营养

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: hashgetdaily_nutrition,
arguments: { date: }
},
id: 1
}

显示：卡路里、蛋白质（克）、碳水化合物（克）、脂肪（克）、纤维（克）、连续天数。

每日仪表盘

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: hashdailydashboard,
arguments: { date: }
},
id: 1
}

查看餐食历史

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: hashgetmeal_history,
arguments: {
limit: 10,
date:
}
},
id: 1
}

对 analysis 字段进行 JSON.parse——请参阅下面的解析餐食分析数据。

删除一餐

步骤 1 —— 获取餐食 ID：
json
{
jsonrpc: 2.0,
method: tools/call,
params: { name: hashgetmeal_history, arguments: { limit: 20 } },
id: 1
}

步骤 2 —— 确认：删除在[时间]记录的[菜品名称]？

步骤 3 —— 删除：
json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: hashdeletemeal,
arguments: { id: }
},
id: 2
}

周报

json
{
jsonrpc: 2.0,
method: tools/call,
params: { name: hashweeklyreport, arguments: {} },
id: 1
}

列出药物

json
{
jsonrpc: 2.0,
method: tools/call,
params: { name: hashgetmedications, arguments: {} },
id: 1
}

添加药物

json
{
jsonrpc: 2.0,
method: tools/call,
params: {
name: hashaddmedication,
arguments: {
name: <药物名称>,
dosage: <例如 500mg>,
frequency: <例如每日两次>,
timeofday: [早晨, 晚上

hash-health哈希健康

CRITICAL RULES

Auth check

Setup

Log a meal — from image

Step 1 — Upload the image immediately (no text to user first)

Step 2 — Call detectIngredients with your visual analysis

Step 3 — Show result and ask for confirmation

Step 4 — Wait for confirmation, then call analyze_food

Log a meal — from text

Check today's nutrition

Daily dashboard

View meal history

Delete a meal

Weekly report

List medications

Add a medication

Delete a medication

Bedtime summary

Nutrition chat

Parsing meal analysis data

Rules

关键规则

身份验证检查

设置

记录一餐——通过图片

步骤 1 —— 立即上传图片（不要先给用户发文字）

步骤 2 —— 使用你的视觉分析调用 detectIngredients

步骤 3 —— 显示结果并请求确认

步骤 4 —— 等待确认，然后调用 analyze_food

记录一餐——通过文本

查看今日营养

每日仪表盘

查看餐食历史

删除一餐

周报

列出药物

添加药物

标签

通过对话安装

方式一：安装 SkillHub 和技能

方式二：设置 SkillHub 为优先技能安装源

通过命令行安装

下载

相关推荐

self-improvement

self-improvement

self-improvement

self-improvement

Step 2 — Call `detectIngredients` with your visual analysis

Step 4 — Wait for confirmation, then call `analyze_food`