tomsbin/mcd-fit-planner

},   "evidence": {     "url": "https://www.mcdonalds.com.cn/product/Corn-Cup/",

"evidence": {     "url": "https://www.mcdonalds.com.cn/product/Corn-Cup/",     "product_name": "玉米杯",

"url": "https://www.mcdonalds.com.cn/product/Corn-Cup/",

<div class="table-tools"><h3>吃麦餐食 · 目标明细</h3><label>筛选 <select id="filter"><option value="all">全部结果</option><option value="within">达标</option><option value="estimated_within">估算达标</option><option value="outside">超出或不足</option><option value=

<h3>全天记录 · 另行核对</h3><p>只有用户确认全天吃喝录全的日期才判断全天目标。新增或纠错会撤销该日确认；一餐达标不等于全天或训练效果达标。</p><div class="scroll"><table><thead><tr><th>日期</th><th>全天录全</th><th>结果</th><th>实际 / 目标</th></tr></thead><tbody>$daily_rows</tbody></table></div> <h3>统计说明</h3><ul>

<h3>统计说明</h3><ul>$notes</ul><p>未标明食物来源的记录 $unclassified 条，未计入吃麦摄入。$privacy_note</p><p><a class="source-link" href="https://www.mcdonalds.com.cn/product/Big-Mac" target="_blank" rel="noopener noreferrer">商品图片来源：麦当劳中国官网 · 巨无霸</a>。商品图仅作视觉配图，不代

Executable file

可直接从克隆的仓库复制 `skill/mcd-fit-planner`。在 Codex 中查看 `/skills` 或用 `$mcd-fit-planner` 明确调用；若尚未发现，重启宿主。路径与发现方式依据 [OpenAI Docs 的技能文档](https://learn.chatgpt.com/docs/build-skills)。

[mcp_servers.mcd] url = "https://mcp.mcd.cn"

url = "https://mcp.mcd.cn"

先在自己的系统／启动环境中设置 `MCD_MCP_TOKEN`，再启动 Codex，使宿主能读取变量。这里只写环境变量名，不写真实 Token。HTTP MCP 与 `bearer_token_env_var` 的含义见 [OpenAI Docs 的 MCP 文档](https://learn.chatgpt.com/docs/extend/mcp?surface=cli)。仓库另有可复制的 `examples/codex-config.example.toml`。

DeepSeek 是模型提供方。需要一个同时支持 DeepSeek 工具调用、MCP 桥接、技能文件和本地 Python 的 Agent 宿主；普通网页聊天框没有这些能力时，不能直接安装或执行本技能。[DeepSeek 官方工具调用文档](https://api-docs.deepseek.com/guides/tool_calls/)说明：模型提出调用请求，具体函数由客户端执行。

[Credential-related source omitted]

[Credential-related source omitted]

端点 https://mcp.mcd.cn ，Streamable HTTP，通过宿主连接器保存认证信息。以服务器实时返回的 tools/list、inputSchema 和实际 response 为准，本文不假定官方 JSON 字段。不要手写未经验证的请求参数，也不要索要或打印明文 Token。

官方能力来源：https://github.com/M-China/mcd-mcp-server/blob/main/README.md

"nutrition": {"energy_kj": 223, "protein_g": 2, "carb_g": 7, "fat_g": 1},     "evidence": {       "url": "https://www.mcdonalds.com.cn/product/Corn-Cup/",

"evidence": {       "url": "https://www.mcdonalds.com.cn/product/Corn-Cup/",       "product_name": "玉米杯", "serving": "小杯，单份", "region": "CN",

"url": "https://www.mcdonalds.com.cn/product/Corn-Cup/",

},     "evidence": {       "url": "https://example.invalid/format-example-only",

"evidence": {       "url": "https://example.invalid/format-example-only",       "product_name": "测试饮品", "serving": "测试标准单杯", "region": "CN",

"url": "https://example.invalid/format-example-only",

人口参考来源：[国家卫健委标准 PDF](https://www.nhc.gov.cn/wjw/yingyang/201710/fdade20feb8144ba921b412944ffb779/files/1739783563332_90405.pdf)。成年人18–49岁轻活动示例男性2250、女性1800kcal/日；真实个体需考虑体重与活动水平。健身总能量采用用户已录入的每日计划，不用上述群体值替用户编造增肌或减脂目标。

参考 [Sam-Hui-dot/mcd-budget-meal](https://github.com/Sam-Hui-dot/mcd-budget-meal/tree/25d5f345862186e9a181339118051cd4418a97e6)（2026-10-24 核查）的预算余量、超预算参考、组成差异及离线报告体验。`meal_report.py`、`assets/meal-report.html` 及推荐引擎的预算参考字段为本项目独立实现，未复制该项目代码、模板

参考 [mai-suanpan](https://github.com/W313997791/mai-suanpan/tree/49cb59a87a421ed9de9a8befe16c1aedc80e41df)（2026-10-20 核查）的套餐与券比较思路、原单与推荐方案对照呈现。`fit_core.py::basket_summary` 为本项目独立实现，未复制该项目代码、提示词、图片或对话；其历史价格、单券限制和实测结论不作为本项目已验证能力。真实优惠以当前门店官方返回

`assets/big-mac-official.png` 原样取自[麦当劳中国官网「巨无霸」产品页](https://www.mcdonalds.com.cn/product/Big-Mac)的透明底商品图，核查日期 2026-10-15。原图地址：<https://officiaweb-afd.mcdonalds.com.cn/public/upload/attachment/2022/08/202208081011205699.png>。仅作商品视觉配图，不代表用户实际

分章节数据叙事参考[网易云音乐 2024 年度听歌报告介绍](https://socialbeta.com/campaign/23649)及[B站 2024 年度报告官方介绍](https://www.bilibili.com/opus/1012957518392983553)。页面、文案规则与排版为本项目实现，未使用这些平台的页面代码、插画或标识。

`mcp_adapter.py::extract_threshold_gap` 改编自 [mcd-coupon-strategist](https://github.com/Ayzzz123/mcd-coupon-strategist/blob/e391fd00ad7322ddddd5d15d88c403c5459f0f5d/src/mcd_adapter.py) 的同名函数，提交 e391fd00ad7322ddddd5d15d88c403c5459f0f5d。

在 https://open.mcd.cn/mcp 登录并激活自己的MCP Token。在WorkBuddy「连接器」→「自定义连接器」→「配置MCP」添加官方Streamable HTTP服务，端点 https://mcp.mcd.cn 。配置结构见仓库根目录mcp-config.example.json。

[Credential-related source omitted]

WorkBuddy技能说明：https://open.workbuddy.cn/docs/skill

麦当劳官方接入说明：https://github.com/M-China/mcd-developer-innovation-challenge

Executable file

args = {k: copy.deepcopy(selection["parameters"][k]) for k in adapter.COUPON_SCOPE_FIELDS if k in selection["parameters"]}         args["code"] = code  # query-meal-detail uses code, not productCode.         requests.append({"request_id": c

args["code"] = code  # query-meal-detail uses code, not productCode.         requests.append({"request_id": code, "tool": "query-meal-detail", "arguments": args})     return {"scope": scope(selection), "requests": requests, "reused": reused

requests.append({"request_id": code, "tool": "query-meal-detail", "arguments": args})

Executable file

[Credential-related source omitted]

Executable file

Accepts official success envelopes or their single JSON MCP text-content wrapper. Field contracts: https://open.mcd.cn/mcp/doc (checked 2026-10-14).

Field contracts: https://open.mcd.cn/mcp/doc (checked 2026-10-14).

blocked_candidates.append({"key": candidate["key"], "reason": str(exc)})             continue         requests.append({"key": candidate["key"], "request_id": candidate["key"], "tool": "calculate-price", "arguments": payload, "variant": "原参数

continue         requests.append({"key": candidate["key"], "request_id": candidate["key"], "tool": "calculate-price", "arguments": payload, "variant": "原参数"})     if isinstance(max_per_candidate, bool) or not isinstance(max_per_candidate, i

requests.append({"key": candidate["key"], "request_id": candidate["key"], "tool": "calculate-price", "arguments": payload, "variant": "原参数"})

def normalize_quotes(requests, responses):     if requests.get("kind") != "official":

if requests.get("kind") != "official":

timestamp = datetime.strptime(response["datetime"], "%Y-%m-%d %H:%M:%S").replace(tzinfo=CN_TIME)             if not 0 <= (datetime.now(CN_TIME) - timestamp).total_seconds() <= 900 or (requests.get("day") and timestamp.date().isoformat() != 

if not 0 <= (datetime.now(CN_TIME) - timestamp).total_seconds() <= 900 or (requests.get("day") and timestamp.date().isoformat() != requests["day"]):

if quote["key"] not in grouped or (quote["total_cents"], quote["request_id"]) < (grouped[quote["key"]]["total_cents"], grouped[quote["key"]]["request_id"]):             grouped[quote["key"]] = quote     coupon_comparison = requests.get("cou

grouped[quote["key"]] = quote     coupon_comparison = requests.get("coupon_comparison", {})     return {"kind": "official", "context_id": requests["context_id"], "store_code": requests["store_code"], "context": requests["context"], "quotes"

coupon_comparison = requests.get("coupon_comparison", {})

Executable file

Executable file

REFERENCE_URL = "https://www.nhc.gov.cn/wjw/yingyang/201710/fdade20feb8144ba921b412944ffb779/files/1739783563332_90405.pdf"

Executable file

Executable file

Executable file

def rule(y, color="#d7d0c4", x=56, end=1024):         add(f'<path d="M{x} {y}H{end}" stroke="{color}"/>')     add(f'<svg xmlns="http://www.w3.org/2000/svg" width="{WIDTH}" height="{HEIGHT}" viewBox="0 0 {WIDTH} {HEIGHT}" role="img" aria-lab

add(f'<path d="M{x} {y}H{end}" stroke="{color}"/>')     add(f'<svg xmlns="http://www.w3.org/2000/svg" width="{WIDTH}" height="{HEIGHT}" viewBox="0 0 {WIDTH} {HEIGHT}" role="img" aria-labelledby="title desc">')     add('<title id="title">吃麦的

add(f'<svg xmlns="http://www.w3.org/2000/svg" width="{WIDTH}" height="{HEIGHT}" viewBox="0 0 {WIDTH} {HEIGHT}" role="img" aria-labelledby="title desc">')

Executable file

Executable file

[Credential-related source omitted]

Path(destination).write_bytes(base64.b64decode(match[1], validate=True))