← 返回 Skills 市场
304
总下载
0
收藏
0
当前安装
10
版本数
在 OpenClaw 中安装
/install nutrition-claw
功能描述
Local CLI for tracking nutrition with JSON data, semantic search, configurable goals, food library, meal logging, impact feedback, and YAML output.
安全使用建议
This package appears to do what it says: a local CLI that stores data under ~/.nutrition-claw and uses a local embedding model for semantic search. Before installing, consider: 1) npm install will pull native/ML libraries (ONNX runtime, transformers, embedding model) and may download large model files on first run—if you want fully offline/no-network behavior, avoid using search features. 2) npm link will create a global CLI symlink and modifies your system Node environment; run in a controlled environment (container or VM) if unsure. 3) The tool writes data to ~/.nutrition-claw (goals.json, foods.json, logs/, vectors/, education.txt); back up or inspect that directory if you care about data placement. 4) AGENTS.md refers to a sync script that would copy SKILL.md into ~/.openclaw/workspace; review or avoid running such scripts unless you trust them. Overall the footprint looks proportionate to the stated purpose; if you want extra assurance, review the package.json/dependencies and run the CLI in an isolated environment before trusting it with real data.
功能分析
Type: OpenClaw Skill
Name: nutrition-claw
Version: 1.0.9
The nutrition-claw skill bundle is a legitimate, well-structured local nutrition tracking CLI. It manages meal logs, a food library, and nutritional goals using local JSON storage in the user's home directory (~/.nutrition-claw/). It utilizes the 'vectra' library and '@themaximalist/embeddings.js' for local semantic search functionality. The extensive instructions in SKILL.md regarding 'educational remarks' and 'feedback diaries' are designed to guide the AI agent in providing nutritional context to the user and do not contain malicious prompt-injection triggers or instructions to exfiltrate data. No suspicious network activity, obfuscation, or unauthorized file access was detected.
能力评估
Purpose & Capability
Name/description (local nutrition CLI, semantic search, YAML output) match the included source: commands for configure/goals/food/meal/summary/history, local JSON storage under ~/.nutrition-claw, and a local vector index for semantic search. Dependencies (embeddings, vectra, transformers) are appropriate for the described semantic search capability.
Instruction Scope
SKILL.md instructs the user to run `npm install`, optional build (Bun), and `npm link` to expose the CLI—these steps are needed to run the packaged Node CLI but modify the developer environment (install packages, create a global symlink). SKILL.md and AGENTS.md also mention downloading a local embedding model on first search and copying SKILL.md into an OpenClaw workspace via a sync script; the latter expects a ./scripts/sync-skill.sh which is referenced but not present. Overall the instructions stay within the tool's purpose but require network/native installs and an explicit sync step that writes to the user's OpenClaw workspace if executed.
Install Mechanism
There is no registry install spec in the skill bundle (instruction-only), but SKILL.md tells users to run `npm install` (pulling dependencies) and `npm link`. Dependencies include native components (ONNX runtime mentioned) and large ML libraries (@xenova/transformers, @themaximalist/embeddings.js), which will cause model & binary downloads at install or first-run. These downloads are expected for local embeddings but are higher-risk than a pure JS-only CLI because they fetch and install native/model artifacts.
Credentials
The skill declares no required environment variables, credentials, or config paths. At runtime it uses only HOME (process.env.HOME) to create ~/.nutrition-claw and reads/writes local files there; that is proportionate to a local CLI storing user data.
Persistence & Privilege
The skill is not always-enabled and does not request elevated privileges. It does perform actions that persist data locally (writes JSON and a vectors dir under ~/.nutrition-claw) and recommends `npm link` (creates a global symlink). The AGENTS.md sync instruction would copy SKILL.md into an OpenClaw workspace if the referenced script is run; this is a normal developer convenience but worth noting because it writes into the agent workspace.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install nutrition-claw - 安装完成后，直接呼叫该 Skill 的名称或使用
/nutrition-claw触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.9
Rename bin/nutrition-claw → bin/nutrition-claw.js so ClawHub's text-file filter includes it; update package.json bin path accordingly
v1.0.8
Add bin/nutrition-claw shim so fresh ClawHub installs can npm link correctly after npm install + npm run build
v1.0.7
nutrition-claw 1.0.7
- Revised food library to support nutrition data entry per arbitrary reference amount, not just per 100g/100ml.
- Updated documentation: food add examples now include various units (g, ml, tbsp), and the "food education" feature is removed.
- Added a new ingredient nutrition commentary: every ingredient addition now explains, in-depth, the nutritional qualities, health impact, misconceptions, and context-specific reasoning behind each food.
- Removed in-app surface of education insights on food add (this is now handled as commentary when adding ingredients).
- Code and documentation updated for clarity and consistency around food units and ingredient feedback.
v1.0.6
Replace JSON education log with a plain text file (education.txt): 20 lines max, one food name per line, FIFO rotation. Keeps token cost minimal while naturally cycling foods back into education eligibility over time.
v1.0.5
Add one-time nutritional education on food add: highlights protein density, calorie density, sat fat, sugar, fiber, and healthy fat ratio vs user goals. Education is stored in ~/.nutrition-claw/education.json and never repeated for the same food.
v1.0.4
Food library now defaults to per 100g/100ml (label-based products only); updated SKILL.md to clarify food library is for packaged products with nutrition labels, not restaurant dishes or home-cooked meals
v1.0.3
- Removed the scripts/sync-skill.sh file.
- No user-facing changes to commands, features, or documentation.
- Internal cleanup: skill now ships without the sync-skill.sh helper script.
v1.0.2
- Added a build step to the setup instructions, including details on building from source using Bun and the "npm run build" command.
- Clarified that the "dist/" folder ships pre-built but can be recompiled if needed.
- Updated step numbering in the setup section for accuracy.
- No changes to functionality or command syntax.
v1.0.1
nutrition-claw 1.0.1
- Initial public release: fully local, offline-first nutrition tracking CLI for openclaw.
- Tracks meals and ingredients, manages a personal food library, and sets daily nutrition goals.
- Implements semantic search (MiniLM, local embeddings) for foods, meals, and ingredients — no network or API keys required.
- All data stored privately in user home directory as plain JSON; model is cached after initial download.
- Commands include: `configure` for goals, `goals`, `food`, `meal`, `summary`, `history`, and advanced search.
- Provides actionable, context-aware nutrition feedback after every ingredient change.
v1.0.0
- Initial release of nutrition-claw: a fully local nutrition tracking CLI with YAML output.
- All user data is stored as local JSON; nothing is sent online—semantic search runs locally.
- CLI provides commands to set nutrition goals (interactive & non-interactive), manage a food library, and log meals with precise date/time.
- Food library supports semantic search, customizable units (weight & volume), and rapid addition from nutrition labels.
- Meal logging includes per-ingredient impact feedback—giving explicit nutritional coaching on goals and overages.
- Includes summary, history, and semantic search functionalities for meals, ingredients, and foods.
元数据
常见问题
Nutrition Claw 是什么？
Local CLI for tracking nutrition with JSON data, semantic search, configurable goals, food library, meal logging, impact feedback, and YAML output. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 304 次。
如何安装 Nutrition Claw？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install nutrition-claw」即可一键安装，无需额外配置。
Nutrition Claw 是免费的吗？
是的，Nutrition Claw 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Nutrition Claw 支持哪些平台？
Nutrition Claw 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Nutrition Claw？
由 Pita（@pita）开发并维护，当前版本 v1.0.9。
推荐 Skills