Token导航 LogoToken导航

Agent Skill · 研究检索

smart-tasks(智能任务)

smart-tasks 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

smart-tasks 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为只读;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 30,718。

总安装 2,832周安装 118GitHub Stars 公开资料未说明下载 944核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:smart-tasks(智能任务)
来源仓库:https://github.com/dgkim311/smart-tasks
安装命令:
openclaw skills install smart-tasks
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install smart-tasks

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

基于 Markdown 文件的智能任务管理系统支持自然语言交互操作。适用于工作区内的优先级设定、提醒设置与每周回顾安排。通过上下文感知实现任务关联与进度自动同步更新。smart-tasks 属于研究检索类 Skill,可作为该场景下的辅助能力补充。在 OpenClaw 中聊天过程中可随时添加、修改或查询任务状态。

使用前请确保工作区目录可读写并配置好任务文件命名规范。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install smart-tasks;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

smart-tasks 怎么安装?

在终端执行 openclaw skills install smart-tasks。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

smart-tasks 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

smart-tasks 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

smart-tasks 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

smart-tasks 用的人多吗?

公开数据是累计安装 2,832 次,近一周安装 118 次,下载 944 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

smart-tasks 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
smart-tasks

智能任务 · 本页

2,832—只读MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
smart-tasks
description
>

Smart Tasks

Manage tasks as markdown files in tasks/ with an INDEX.md summary. Minimize context usage: read INDEX.md first, individual files only when needed.

Setup

Run the init script once to create the directory structure:

bash skills/smart-tasks/scripts/init-tasks.sh

This creates tasks/, tasks/active/, tasks/done/, tasks/archive/, tasks/INDEX.md, and tasks/.meta.json. Idempotent — safe to re-run.

After init, apply workspace integration changes from workspace-integration.md with user approval.

File Structure Overview

  • tasks/INDEX.md — active task summary table (~1-2KB). Read every session.
  • tasks/.meta.json — tracks nextId counter and known categories.
  • tasks/active/ — individual task files (one per task).
  • tasks/done/ — completed tasks (kept 30 days).
  • tasks/archive/YYYY-QN/ — long-term storage.

For full format specs see file-structure.md.

Core Workflow

1. Detect Task Intent

When the user mentions deadlines, todos, or things to do in conversation, propose creating a task. No special commands needed — use natural language.

Examples of triggers:

  • "다음 주 금요일까지 리뷰 완료해야 해"
  • "논문 피드백 보내야 하는데"
  • "이번 주 할일 뭐 있어?"

2. Create a Task

  1. Read tasks/.meta.json to get nextId.
  2. Generate ID: T-{nextId} zero-padded to 3 digits (e.g., T-001).
  3. Create slug from title (kebab-case English summary).
  4. Write task file to tasks/active/{ID}_{slug}.md using the format in

task-format.md.

  1. Fill the Agent Notes section with relevant workspace context

(related files, people, timeline analysis).

  1. Increment nextId in .meta.json. Add new category if not already listed.
  2. Update tasks/INDEX.md — insert row in the correct section.

3. Update a Task

  1. Identify the task by ID or title search in INDEX.md.
  2. Read the individual task file from tasks/active/.
  3. Apply changes to both the task file and INDEX.md.
  4. Append a timestamped entry to the task's 메모/진행 기록 section.

4. Complete a Task

  1. Set status: done and completed: {today} in the task file.
  2. Move file from tasks/active/ to tasks/done/.
  3. Remove the row from INDEX.md. (Do NOT add to done section in INDEX.)
  4. Confirm completion to the user.

5. Query Tasks

  • Quick overview: Read INDEX.md only. Answer from the table.
  • Detail needed: Read the specific task file(s) from tasks/active/.
  • "What should I do?": Analyze INDEX.md by deadline × priority × estimated hours.

Recommend a sequence with reasoning.

6. Delete a Task

Only when explicitly requested. Move to tasks/done/ with status: cancelled rather than deleting the file. Remove from INDEX.md.

INDEX.md Management

INDEX.md is the single-read entry point for task awareness.

Structure

# Tasks Index

> Last updated: YYYY-MM-DD
> Active: N | Overdue: N | Due this week: N

## 🔴 Overdue
| ID | Title | Due | Priority | Category |
|----|-------|-----|----------|----------|

## 🟡 Due This Week
| ID | Title | Due | Priority | Category |
|----|-------|-----|----------|----------|

## 📋 Later
| ID | Title | Due | Priority | Category |
|----|-------|-----|----------|----------|

## 📊 Summary

Update Rules

  • Update INDEX.md on every task create / update / complete / delete.
  • Recalculate header counters (Active, Overdue, Due this week) each update.
  • Keep rows sorted by due date within each section.
  • Reclassify rows between sections based on current date.
  • Set Last updated to today.

Size Control

Active tasks~SizeAction
1–20~1KBKeep as-is
20–50~2.5KBAbbreviate Later section (title + due only)
50+~5KB+Split into category indexes (INDEX-{category}.md)

Context Strategy

Read context-strategy.md for details on:

  • Layered reading pattern (INDEX → active file → done/archive)
  • Subagent cost optimization for cron jobs
  • When to read vs. skip individual files

Key rule: If INDEX.md alone can answer the question, do NOT read individual files.

Cron Jobs

Set up scheduled briefings and alerts. All cron jobs use Sonnet-class models. See cron-templates.md for:

  • Daily morning briefing (08:00) — overdue alerts + today's priorities
  • Deadline check (12:00, 18:00) — D-0/D-1/overdue only, silent if none
  • Weekly review (Sunday 20:00) — completion summary, stale task detection,

archive cleanup (done/ 30+ days → archive/)

Automatic Cleanup

TransitionTriggerAction
active → doneUser marks completeMove file, update INDEX
done → archive30 days elapsedWeekly review moves to archive/YYYY-QN/
archive—Never in INDEX. Access only via search.

Task File Format

Each task is a markdown file with YAML frontmatter. See task-format.md for:

  • Complete frontmatter schema
  • Section template
  • 2-3 example tasks

Workspace Integration

This skill does NOT modify files outside its own directory. See workspace-integration.md for recommended changes to AGENTS.md, SOUL.md, and HEARTBEAT.md (diff format). Apply these with user approval during setup.

Integration with Other Skills

  • proactivity: Record task status in session-state; proactively surface

relevant tasks during conversation.

  • self-improving: Log task management mistakes/improvements to domain file.

Create self-improving/tasks.md for task-specific learnings.

Best Practices

  1. Always update INDEX.md and the task file atomically.
  2. Fill Agent Notes with cross-references to workspace files at creation time.
  3. Use estimated_hours to give realistic scheduling advice.
  4. When reminding about tasks, include relevant context — not just the deadline.
  5. Categories are user-defined. Never hardcode a fixed set.
  6. Keep task titles concise but descriptive for INDEX readability.

平台分布

OpenClaw

98.3%
按下载量换算928

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills