Midlong Term Task Manager

管理和跟踪 AI Agent 中长期任务，支持任务分解、进度更新、延期提醒及每日/每周总结生成。

Security Scan

The skill is internally coherent: it provides local shell scripts to manage task JSON and logs under your ~/.openclaw workspace, does not request credentials, does not contact external endpoints, and matches its description.

Purpose & Capability

Name/description (mid/long-term task manager) align with the provided artifacts: SKILL.md plus three shell scripts that read/write a tasks.json and logs under ~/.openclaw/workspace/.tasks. There are no unrelated env vars, binaries, or network endpoints requested.

Instruction Scope

Instructions and SKILL.md stay within scope: they instruct running the included scripts, adding heartbeat/cron entries, and storing data under ./.openclaw/workspace/.tasks. Note: the scripts persist task data and logs to disk (tasks.json, logs, summaries/weekly reports); the user should be aware that these files are stored locally and accessible to local processes and users with filesystem access.

Install Mechanism

No install spec (instruction-only + bundled scripts). Scripts are plain shell files included in the package — no network downloads, no package installers, and no extracted archives. This is low-risk from an install mechanism perspective.

Credentials

Skill declares no required environment variables or credentials and the code does not attempt to read secrets or unrelated config. It optionally uses jq if installed, which is reasonable for JSON parsing. The number and type of accesses are proportionate to a local task-manager skill.

Persistence & Privilege

always is false and the skill does not request permanent platform-level privileges or modify other skills' configs. It only writes its own task/log files under ~/.openclaw/workspace/.tasks and suggests (user-driven) cron/heartbeat integration.

Assessment

This skill appears safe and coherent for local task tracking, but review before enabling: 1) inspect the included scripts (they are small, readable, and local-only) and confirm you are comfortable with files being created at ~/.openclaw/workspace/.tasks (tasks.json, logs, summaries). 2) Ensure tasks.json will not contain sensitive secrets; store only task metadata. 3) If you enable cron/heartbeat entries, add them manually rather than blindly executing instructions. 4) Consider file permissions (chmod) on the .tasks directory to limit access. 5) The scripts optionally use jq and rely on GNU date extensions on some platforms — test on your OS or adjust for portability.

中长期任务管理技能

技能说明

名称： midlong-term-task-manager
作用： 管理 AI Agent 的中长期任务（几天到几个月），支持任务分解、进度跟踪、自动提醒

与三层记忆系统的边界：

• ✅ 任务数据存储在独立文件：.tasks/tasks.json
• ✅ 进度日志记录到：.tasks/logs/YYYY-MM-DD.md
• ❌ 不直接编辑 MEMORY.md（由三层记忆技能自动压缩）

---

触发场景

1. 心跳检查时 - 自动 review 所有进行中的任务
2. 用户添加任务 - 登记新任务并分解
3. 任务状态变更 - 更新进度/延期/完成
4. 每日总结 - 生成任务进度报告

---

任务数据结构

{
  "tasks": [
    {
      "id": "TSK-20260330-001",
      "name": "学习中长期任务管理",
      "description": "悠悠学习任务管理技能的开发",
      "type": "learning",
      "priority": "high",
      "status": "in_progress",
      "created_at": "2026-03-30T10:30:00+08:00",
      "due_date": "2026-04-06",
      "decomposed": [
        {
          "subtask": "设计数据结构",
          "status": "done",
          "due_date": "2026-03-30"
        },
        {
          "subtask": "编写 SKILL.md",
          "status": "in_progress",
          "due_date": "2026-03-30"
        },
        {
          "subtask": "创建心跳检查脚本",
          "status": "pending",
          "due_date": "2026-03-31"
        }
      ],
      "progress": 33,
      "last_updated": "2026-03-30T10:30:00+08:00",
      "blocked_by": null,
      "depends_on": [],
      "tags": ["skill-development", "self-improvement"]
    }
  ]
}

---

任务状态

状态	说明	心跳检查动作
pending	待开始	检查是否到达开始日期
in_progress	进行中	记录进度，检查是否延期
blocked	阻塞	提醒解除阻塞条件
completed	已完成	归档，生成总结
paused	暂停	记录暂停原因
cancelled	已取消	归档

---

使用方法

添加任务

# 命令格式
task_add <name> --due <YYYY-MM-DD> --priority <low|medium|high> --desc <description>

# 示例
task_add "学习具身智能" --due 2026-04-15 --priority high --desc "帮助智哥了解具身智能基础"

更新进度

# 命令格式
task_update <task_id> --progress <0-100> --status <status> --note <note>

# 示例
task_update TSK-20260330-001 --progress 50 --status in_progress --note "完成 SKILL.md"

查询任务

# 查看所有任务
task_list

# 查看进行中的任务
task_list --status in_progress

# 查看即将到期的任务（7 天内）
task_list --due-soon 7

删除任务

task_cancel <task_id> --reason <reason>

---

心跳检查集成

在 HEARTBEAT.md 中添加：

### 【6】任务进度检查
- 运行 `~/.openclaw/skills/midlong-term-task-manager/scripts/task_check.sh`
- 检查所有 `in_progress` 任务
- 记录进度到 `.tasks/logs/YYYY-MM-DD.md`
- 如有延期风险，提醒悠悠

---

Cron 配置

# 任务检查（心跳时运行，每 30 分钟）
# 在 HEARTBEAT.md 中调用脚本

# 每日总结（每天 20:00）
0 20 * * * ~/.openclaw/skills/midlong-term-task-manager/scripts/daily_summary.sh >> ~/.openclaw/workspace/.tasks/daily.log 2>&1

# 每周 review（每周一 09:00）
0 9 * * 1 ~/.openclaw/skills/midlong-term-task-manager/scripts/weekly_review.sh >> ~/.openclaw/workspace/.tasks/weekly.log 2>&1

---

与三层记忆系统集成

正确做法：

1. ✅ 任务进度日志 → .tasks/logs/YYYY-MM-DD.md
2. ✅ 任务完成总结 → .learnings/LEARNINGS.md
3. ✅ 由三层记忆技能自动压缩到 MEMORY.md

错误做法：

• ❌ 直接编辑 MEMORY.md
• ❌ 直接写入 memory/YYYY-MM-DD.md（这是三层记忆技能的 L2 存储）

---

文件结构

~/.openclaw/skills/midlong-term-task-manager/
├── SKILL.md              # 技能说明
└── scripts/
    ├── task_check.sh     # 心跳检查脚本
    ├── daily_summary.sh  # 每日总结脚本
    └── weekly_review.sh  # 每周 review 脚本

~/.openclaw/workspace/.tasks/
├── tasks.json            # 任务数据库
└── logs/
    └── YYYY-MM-DD.md     # 每日进度日志

---

最佳实践

1. 及时登记 - 接到任务立即记录，避免遗忘
2. 合理分解 - 长期任务分解为周/日计划
3. 定期 review - 心跳时自动检查进度
4. 诚实记录 - 延期不隐瞒，及时调整计划
5. 完成总结 - 每个任务完成后记录学习心得

---

版本历史

• v0.1.0 (2026-03-30) - 初始版本，基础功能