menus-recommender — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited menus-recommender (Plugin) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
一个 AI agent 插件：当用户表达就餐意向（「推荐晚餐」「不知道吃什么」「想吃点辣的」），基于 462 道菜的全球美食知识库做加权随机推荐。
python3 scripts/recommender.py --count 3 --keywords "辣,牛肉"返回 JSON：
{"dishes": [{"id": "...", "name": "麻婆豆腐", "price": 38, "cuisine": "川菜", ...}, ...], "exhausted": false, ...}完整参数：
python3 scripts/recommender.py --help方式 A：自建 marketplace（推荐）
/plugin marketplace add spyyps/recommend-dish
/plugin install menus-recommender@spyyps-recommend-dish安装后在任意目录的 Claude Code 会话里说「推荐晚餐」「想吃点辣的」即可触发。 后续更新插件：/plugin update menus-recommender。
方式 B：本地 git clone
git clone https://github.com/spyyps/recommend-dish ~/.claude/plugins/menus-recommendergit clone https://github.com/spyyps/recommend-dish
cd recommend-dish
gemini # 在本目录运行 Gemini CLI，会自动加载 GEMINI.md待添加（参考 .cursor-plugin/ 与 .codex-plugin/ 的多 manifest 适配，第二阶段交付）。
适用于 Claude Desktop / Cline / Cursor / Continue 等任何 MCP 客户端。
uvx 方式（推荐，无需 pip install）：
{
"mcpServers": {
"menus": {
"command": "uvx",
"args": ["menus-mcp"]
}
}
}或者 pip install：
pip install menus-mcp{
"mcpServers": {
"menus": { "command": "menus-mcp" }
}
}| 用户说 | 触发 | 解析为 |
|---|---|---|
| 「推荐晚餐」 | ✓ | 默认 3 道，无筛选 |
| 「想吃点辣的」 | ✓ | --keywords "辣" |
| 「来 5 道便宜的海鲜」 | ✓ | --count 5 --keywords "海鲜" --price-tier "实惠" |
| 「推荐两道川菜」 | ✓ | --count 2 --cuisine "川菜" |
| 「换一批」（紧接上一次推荐） | ✓ | --exclude-ids "<上轮所有 id>" |
| 「100 块以内的欧洲菜」 | ✓ | --max-price 100 --geo "欧洲" |
只接受以下关键词，超出范围的（如「不辣」「清淡」）由 LLM 在调用前消化掉：
menu.json 与 knowledge_base/ 各索引文件git push 到 GitHub（本仓库已就绪：spyyps/recommend-dish）.claude-plugin/marketplace.json 已配置好，列出所有可装插件仓库内 mcp-server/ 目录已实现 MCP server（Python，复用同一 recommender.py），暴露 recommend_dishes 与 list_keywords 两个工具。发布方式：
cd mcp-server
pip install build twine
python3 -m build
twine upload dist/*用户 uvx menus-mcp 或 pip install menus-mcp 即可。
github.com/modelcontextprotocol/servers 把 menus-mcp 加进列表。npx @anthropic-ai/mcpb pack → GitHub Release → 用户拖入 Claude Desktop 即装。MIT
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.