什么是 ClawKeeper？

ClawKeeper 是一款注重隐私的命令行界面，旨在弥合 AI 智能体与个人任务管理之间的鸿沟。通过将所有数据存储在本地机器上的纯 Markdown 文件中，它确保您的生产力数据始终处于您的控制之下。该工具是用户使用 Openclaw Skills 构建自定义工作流的核心组件，提供了一种无需依赖第三方云服务即可处理待办事项列表和习惯养成的原生方式。

该技能使 AI 智能体能够充当支持性的协作伙伴。通过与 CLI 交互，智能体可以维护任务列表、追踪习惯间隔，甚至根据您的活动模式提供主动提醒。由于它通过环境变量依赖共享目录，多个智能体可以使用相同的底层数据源协同工作。

下载入口:https://github.com/openclaw/skills/tree/main/skills/tallhamn/clawkeeper

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。

ClawKeeper 应用场景

• 通过 AI 辅助界面追踪专业项目里程碑和子任务。
• 通过自动智能体提醒，维持锻炼或冥想等日常个人习惯。
• 使用共享的本地 Markdown 仓库在不同的 AI 智能体之间同步任务列表。
• 存档有关任务进度的反思和笔记，用于长期生产力分析。

1. 该技能通过 CLAWKEEPER_DIR 环境变量识别存储路径，以定位 Markdown 数据文件。
2. 用户或智能体发出 CLI 命令，使用 clawkeeper 二进制文件添加、编辑或列出任务和习惯。
3. CLI 处理请求并返回一个表示成功或失败的结构化 JSON 对象。
4. 在定期的心跳检查期间，智能体审查任务和习惯的状态，以识别错过的截止日期或正在形成的连续记录。
5. 智能体根据用户交互和完成状态提供反馈或更新 Markdown 文件。

ClawKeeper 配置指南

要将此工具集成到您的 Openclaw Skills 环境中，请通过 npm 安装软件包并定义您的存储目录。

npm install -g clawkeeper
# 在 shell 配置中设置数据目录
export CLAWKEEPER_DIR="~/.clawkeeper/"

ClawKeeper 数据架构与分类体系

ClawKeeper 使用结构化方法管理信息，确保 Openclaw Skills 能够轻松解析和操作数据。所有输出均以 JSON 对象形式返回，以确保高可靠性。

name: clawkeeper
description: Tasks and habits that live in a plain markdown file on your machine. Free, private, and claw-native.
metadata: {"openclaw": {"requires": {"bins": ["clawkeeper"], "env": ["CLAWKEEPER_DIR"]}, "primaryEnv": "CLAWKEEPER_DIR", "install": [{"id": "npm", "kind": "node", "package": "clawkeeper", "bins": ["clawkeeper"], "label": "Install via npm"}]}}

ClawKeeper CLI

Manage the user's tasks and habits via the ClawKeeper CLI. All data is stored as markdown at the path set by CLAWKEEPER_DIR (defaults to ~/.clawkeeper/).

clawkeeper   [--flags]

If CLAWKEEPER_DIR is set in your environment, the CLI reads and writes data there. This allows multiple agents to share the same task list.

Tasks

clawkeeper task list
clawkeeper task add --text "Buy groceries"
clawkeeper task add --text "Buy groceries" --due-date 2026-03-15
clawkeeper task add-subtask --parent-text "Buy groceries" --text "Milk"
clawkeeper task complete --id 
clawkeeper task complete --text "Buy groceries"
clawkeeper task uncomplete --id 
clawkeeper task edit --text "Old name" --new-text "New name"
clawkeeper task edit --text "Old name" --due-date 2026-04-01
clawkeeper task set-due-date --text "Buy groceries" --due-date 2026-03-15
clawkeeper task set-due-date --text "Buy groceries" --due-date none
clawkeeper task delete --text "Buy groceries"
clawkeeper task add-note --text "Buy groceries" --note "Check prices first"
clawkeeper task edit-note --text "Buy groceries" --note "Check prices first" --new-note "Compare at two stores"
clawkeeper task delete-note --text "Buy groceries" --note "Check prices first"

Habits

clawkeeper habit list
clawkeeper habit add --text "Meditate" --interval 24
clawkeeper habit edit --text "Meditate" --new-text "Morning meditation" --interval 12
clawkeeper habit delete --text "Meditate"
clawkeeper habit complete --text "Meditate"
clawkeeper habit add-note --text "Meditate" --note "Felt calm today"
clawkeeper habit edit-note --text "Meditate" --note "Felt calm today" --new-note "Felt calm, 10 min session"
clawkeeper habit delete-note --text "Meditate" --note "Felt calm today"

State

clawkeeper state show

Proactive Checks (Heartbeat)

When running periodic checks, use clawkeeper state show to review the user's habits and tasks:

• Missed habits: If a habit hasn't been completed for more than 2x its interval, ask about it gently. People forget — a nudge helps more than a lecture.
• Building streaks: When a habit's completion count is climbing, acknowledge the momentum briefly.
• Recent notes: If the user added reflections or notes recently, reference them for continuity. It shows you're paying attention.
• Stale tasks: Tasks sitting open for a long time might need to be broken down, re-prioritized, or dropped.

Tone: supportive collaborator, not drill sergeant. If nothing needs attention, reply HEARTBEAT_OK.

Notes

• All commands return JSON: {"ok": true, "data": ...} on success, {"ok": false, "error": "..."} on failure.
• IDs are stable across invocations. Use --id for precise lookups or --text for fuzzy substring matching.
• When adding a task, the response includes the new task's id for subsequent operations.