← 返回 Skills 市场
auto-complex-task-planner
作者
sixtysixsone
· GitHub ↗
· v2.0.0
· MIT-0
609
总下载
0
收藏
6
当前安装
1
版本数
在 OpenClaw 中安装
/install auto-complex-task-planner
功能描述
自动复杂任务规划器。智能判断任务复杂度，自动创建子 agent 执行，支持优先级队列、任务模板、进度追踪、增强质量审核。 使用场景： - "帮我调研 XXX"（复杂调研任务） - "开发一个 XXX 功能"（开发任务） - "批量处理 XXX"（批量操作） - "生成 XXX 报告/文档"（文档生成） - "紧急...
使用说明 (SKILL.md)
自动复杂任务规划技能
版本: v2.0.0（已实现所有改进建议）
作者: Johnny Liu
创建时间: 2026-03-08
🚀 核心功能
1. 智能任务分析
- ✅ 自动判断任务复杂度（长度、步骤、关键词）
- ✅ 自动识别任务类型（调研/开发/文档/批量）
- ✅ 新增 自动检测优先级（高/普通/低）
- ✅ 估算执行时间
2. 并行子 agent 执行
- ✅ 自动生成任务描述
- ✅ 新增 使用任务模板系统
- ✅ 并行执行独立任务（效率提升 60-80%）
- ✅ 新增 支持优先级队列
3. 增强质量审核
- ✅ 新增 交付物存在性检查
- ✅ 新增 内容质量评估
- ✅ 新增 幻觉检测
- ✅ 多维度质量打分
4. 任务生命周期管理
- ✅ 新增 JSON 格式任务记录
- ✅ 新增 任务进度追踪
- ✅ 新增 每日统计报表
- ✅ 自动清理已完成
5. 任务模板系统
- ✅ 新增 调研任务模板
- ✅ 新增 开发任务模板
- ✅ 新增 文档生成模板
- ✅ 新增 批量处理模板
📋 快速开始
安装
npx clawhub install auto-complex-task-planner
使用示例
用户：紧急！帮我调研北京新发地农产品市场
AI: 检测到高优先级调研任务，将优先执行...
✓ 任务 ID: a1b2c3d4
✓ 优先级：高
✓ 创建 3 个子 agent（并行执行）
✓ 预计完成时间：10-15 分钟
🎯 任务分类
自动使用子 agent
| 类型 | 关键词 | 示例 |
|---|---|---|
| 调研 | 调研、搜集、分析、研究 | "调研 XXX 市场" |
| 开发 | 开发、编写、创建、实现 | "开发 XXX 功能" |
| 批量 | 批量、全部、所有、删除 | "批量删除 XXX" |
| 文档 | 报告、文档、方案、总结 | "生成 XXX 报告" |
优先级识别
| 优先级 | 关键词 | 处理 |
|---|---|---|
| 高 | 紧急、优先、马上 | 优先执行 |
| 普通 | 无 | 正常执行 |
| 低 | 有空、不急 | 空闲时执行 |
📊 效率提升
| 方式 | 耗时 | 效率 |
|---|---|---|
| 主 session 串行 | 45-60 分钟 | 100% |
| 本技能并行 | 10-15 分钟 | 提升 75-83% |
⚙️ 配置
# 环境变量
export SUBAGENT_TIMEOUT=600
export MAX_RETRIES=3
export QUALITY_THRESHOLD=0.7
export MAX_CONCURRENT=5
📁 文件说明
SKILL.md- 技能文档README.md- 英文指南README_CN.md- 中文指南scheduler.py- 执行脚本package.json- 配置
📜 版本历史
v2.0.0 (2026-03-09)
- 新增优先级队列系统
- 新增任务模板系统
- 新增 JSON 任务记录
- 新增进度追踪
- 增强质量审核
v1.0.0 (2026-03-08)
- 初始版本
作者: Johnny Liu
许可: MIT
安全使用建议
Before installing, review the complete scheduler.py (the packaged file shown here is truncated) to ensure it contains no network exfiltration, subprocess execution, or reading of unexpected system files. Ask the author why package.json is referenced but missing from the manifest and why SKILL.md lists env vars that metadata does not declare. If you proceed, run the skill in a restricted sandbox or container, and audit file writes under /home/admin/.openclaw and any outgoing network activity; limit its filesystem permissions and do not run it on a machine containing sensitive logs or credentials until you confirm the code is safe.
功能分析
Type: OpenClaw Skill
Name: auto-complex-task-planner
Version: 2.0.0
The skill is a task management and orchestration utility designed to decompose complex user requests into sub-tasks for parallel execution. The logic in `scheduler.py` uses keyword analysis and predefined templates to categorize tasks (e.g., research, development) and manages their lifecycle via local JSON and Markdown logs in the `/home/admin/.openclaw/workspace/` directory. No evidence of data exfiltration, unauthorized shell execution, or malicious prompt injection was found; the code relies entirely on standard libraries and follows its stated purpose.
能力评估
Purpose & Capability
Name/description describe an agent-based task scheduler and the included scheduler.py and templates implement scheduling, templates, priority queue and file-based task records — this is coherent. However SKILL.md and README refer to a package.json and additional env vars that are not declared in the metadata/manifest (manifest lists 4 files; package.json is referenced but not present), which is an inconsistency worth questioning.
Instruction Scope
SKILL.md instructs or implies reading/writing files under /home/admin/.openclaw (logs, workspace, temp), running/inspecting subagents (e.g., 'subagents list'), and saving outputs to absolute paths. Those file and log accesses are plausible for a scheduler but broaden the skill's scope to filesystem access and potential sensitive log reading; the instructions also reference environment variables and system commands not declared in metadata. The provided scheduler.py (truncated in the package listing presented here) contains hardcoded paths into /home/admin/.openclaw which means the skill will read/write host files — the full source must be reviewed for any network calls, subprocess execution, or credentials access.
Install Mechanism
There is no install spec (instruction-only install via npx clawhub is documented), and no external download URLs or extract operations in the package metadata. This is lower risk than a remote binary download. The skill does include a Python script shipped with the package which will be executed by the agent runtime — review of that script is necessary.
Credentials
The skill metadata declares no required env vars or credentials, but SKILL.md and READMEs list several environment variables (SUBAGENT_TIMEOUT, MAX_RETRIES, QUALITY_THRESHOLD, ENABLE_PARALLEL, ENABLE_PRIORITY, MAX_CONCURRENT). This mismatch (no env requirements declared vs. docs showing env vars) is an inconsistency. The skill does not request cloud/API credentials, which is proportionate to its purpose, but the documented env vars and hardcoded filesystem paths increase its ability to affect or read local host state.
Persistence & Privilege
The skill does not request always:true, does not declare system-wide config modifications, and appears to operate within the user's OpenClaw workspace (/home/admin/.openclaw). That scope is expected for a scheduler; there is no evidence it attempts to persist beyond its own workspace. Still, hardcoded absolute paths mean it will create/read files under that home path.
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 auto-complex-task-planner - 安装完成后，直接呼叫该 Skill 的名称或使用
/auto-complex-task-planner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
v2.0.0
### 1. Intelligent Task Analysis
- ✅ Automatically judge task complexity
- ✅ Identify task type (research/development/documentation/batch)
- ✅ **NEW** Auto-detect priority (high/normal/low)
- ✅ Estimate execution time
### 2. Parallel Sub-Agent Execution
- ✅ Auto-generate task descriptions
- ✅ **NEW** Use task template system
- ✅ Execute independent tasks in parallel (60-80% efficiency boost)
- ✅ **NEW** Support priority queue
### 3. Enhanced Quality Review
- ✅ **NEW** Deliverable existence check
- ✅ **NEW** Content quality assessment
- ✅ **NEW** Hallucination detection
- ✅ Multi-dimensional quality scoring
### 4. Task Lifecycle Management
- ✅ **NEW** JSON format task recording
- ✅ **NEW** Task progress tracking
- ✅ **NEW** Daily statistics report
- ✅ Auto-cleanup completed tasks
### 5. Task Template System
- ✅ **NEW** Research task template
- ✅ **NEW** Development task template
- ✅ **NEW** Documentation template
- ✅ **NEW** Batch processing template
元数据
常见问题
auto-complex-task-planner 是什么？
自动复杂任务规划器。智能判断任务复杂度，自动创建子 agent 执行，支持优先级队列、任务模板、进度追踪、增强质量审核。 使用场景： - "帮我调研 XXX"（复杂调研任务） - "开发一个 XXX 功能"（开发任务） - "批量处理 XXX"（批量操作） - "生成 XXX 报告/文档"（文档生成） - "紧急... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 609 次。
如何安装 auto-complex-task-planner？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install auto-complex-task-planner」即可一键安装，无需额外配置。
auto-complex-task-planner 是免费的吗？
是的，auto-complex-task-planner 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
auto-complex-task-planner 支持哪些平台？
auto-complex-task-planner 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 auto-complex-task-planner？
由 sixtysixsone（@sixtysixsone）开发并维护，当前版本 v2.0.0。
推荐 Skills