什么是 whoop-cli？

whoop-cli 技能是 @andreasnlarsen/whoop-cli 软件包的专用配套工具，旨在为 AI Agent 提供稳定、JSON 格式的 WHOOP 健康与健身数据访问。通过消除对原始 API 的对接需求，该技能允许 Agent 在编程或自动化环境中直接分析恢复、压力、睡眠和健康标记。

利用 Openclaw Skills 进行生物识别数据分析，确保开发者可以构建尊重安全边界的个性化健康追踪 Agent。该技能强调本地身份验证和只读操作命令，在确保敏感凭据安全的同时，为教练、趋势分析和活动解读提供高保真数据。

下载入口:https://github.com/openclaw/skills/tree/main/skills/andreasnlarsen/whoop-cli

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

whoop-cli 应用场景

• 监控每日恢复和压力分数，以提供上下文相关的生产力建议。
• 分析 7 天或 30 天的健康趋势，以识别生理标记或表现下滑。
• 切分活动数据，以区分有意识的训练负荷和偶然的活动。
• 通过结构化的生命周期管理来计划和跟踪健康实验（例如：咖啡因摄入时机或睡眠协议）。
• 将历史健康数据导出到本地 JSONL 文件，用于自定义数据科学或可视化项目。

1. Agent 通过 whoop auth status 命令检查本地身份验证状态来初始化连接。
2. 在确认凭据有效后，Agent 可以执行高级摘要或每日简报命令来检索用户当前的生理状态。
3. 数据以结构化 JSON 格式返回，通常使用 --pretty 标志以提高可读性，或通过 jq 进行特定字段提取。
4. 对于活动分析，Agent 应用 --labeled-only 等过滤器，以确保其仅解释已确认的训练数据。
5. 如果用户发起实验，Agent 会更新存储在本地 experiments.json 文件中的规范状态，随着时间的推移跟踪假设、协议和结果。

whoop-cli 配置指南

要开始使用，请通过 npm 全局安装 whoop-cli 软件包：

npm install -g @andreasnlarsen/whoop-cli@0.3.1

接下来，安装特定的 Openclaw Skills 软件包：

whoop openclaw install-skill --force

您还必须配置以下环境变量：WHOOP_CLIENT_ID、WHOOP_CLIENT_SECRET 和 WHOOP_REDIRECT_URI。请注意，初始登录应由用户在本地 shell 中执行，以保护 OAuth 令牌的安全。

whoop-cli 数据架构与分类体系

该技能跨本地配置文件和结构化命令输出管理数据：

name: whoop-cli
description: Companion skill for @andreasnlarsen/whoop-cli: agent-friendly WHOOP access via stable CLI JSON (day briefs, health flags, trends, exports) without raw API plumbing.
metadata:
  openclaw:
    requires:
      bins:
        - whoop
      env:
        - WHOOP_CLIENT_ID
        - WHOOP_CLIENT_SECRET
        - WHOOP_REDIRECT_URI
    primaryEnv: WHOOP_CLIENT_SECRET
    homepage: https://github.com/andreasnlarsen/whoop-cli
    install:
      - kind: node
        package: "@andreasnlarsen/whoop-cli@0.3.1"
        bins:
          - whoop
        label: Install whoop-cli from npm

whoop-cli

Use the installed whoop command.

Security + credential handling (required)

• Never ask users to paste client secrets/tokens into chat.
• For first-time auth, the user should run login locally on their own shell.
• Prefer read-only operational commands in agent flows (summary, day-brief, health, trend, sync pull).
• Do not run whoop auth login unless the user explicitly asks for login help.
• Tokens are stored locally at ~/.whoop-cli/profiles/.json by the CLI.

Install / bootstrap

If whoop is missing:

npm install -g @andreasnlarsen/whoop-cli@0.3.1

Optional OpenClaw skill install from package bundle:

whoop openclaw install-skill --force

Core checks

1. whoop auth status --json
2. If unauthenticated, ask the user to run local login:
   • whoop auth login --client-id ... --client-secret ... --redirect-uri ...
3. Validate:
   • whoop day-brief --json --pretty

Useful commands

• Daily:
  • whoop summary --json --pretty
  • whoop day-brief --json --pretty
  • whoop strain-plan --json --pretty
  • whoop health flags --days 7 --json --pretty
• Activity analysis:
  • whoop activity list --days 30 --json --pretty
  • whoop activity trend --days 30 --json --pretty
  • whoop activity types --days 30 --json --pretty
  • training-only: whoop activity trend --days 30 --labeled-only --json --pretty

Activity interpretation guardrail (important)

• WHOOP generic activity rows (often sport_id=-1) are auto-detected and may be unlabeled movement (housework/incidental activity), not intentional training.
• Do not treat generic activity as confirmed training volume by default.
• For coaching/training recommendations, default to --labeled-only and report both total vs filtered counts.

Agent filtering pattern (jq-friendly)

• Canonical source: whoop activity list --json
• Prefer built-in filters first (--labeled-only, --generic-only, --sport-id, --sport-name).
• If custom slicing is needed and jq is available, filter shell-side from raw JSON (example):

whoop activity list --days 30 --json | jq '.data.records | map(select(.sport_id != -1))'

• Export:
  • whoop sync pull --start YYYY-MM-DD --end YYYY-MM-DD --out ./whoop.jsonl --json --pretty

Experiment protocol (agent-required)

• Canonical state: ~/.whoop-cli/experiments.json only.
• Plan experiments with context at creation time:
  • whoop experiment plan --name ... --behavior ... --start-date YYYY-MM-DD [--end-date YYYY-MM-DD] --description ... --why ... --hypothesis ... --success-criteria ... --protocol ... --json --pretty
• Update context without creating duplicate state:
  • whoop experiment context --id ... [--description ... --why ... --hypothesis ... --success-criteria ... --protocol ...] --json --pretty
• Check lifecycle/status with:
  • whoop experiment status [--status planned|running|completed] [--id ...] --json --pretty
• Evaluate outcomes with:
  • whoop experiment report --id ... --json --pretty
• Profile scope is strict by default (active --profile only).
  • Use --all-profiles only when cross-profile visibility is explicitly needed.
• Prefer output field sourceOfTruth (path to canonical state file); experimentsFile is kept as compatibility alias.
• Avoid duplicating experiment state into other files unless the user explicitly asks for separate notes.

Safety

• Never print client secrets or raw tokens.
• Keep API errors concise and actionable.
• Treat this integration as unofficial/non-affiliated.