← 返回 Skills 市场
Goal Decomposer
作者
Johnnie23919
· GitHub ↗
· v1.0.0
· MIT-0
58
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install goal-decomposer
功能描述
将高层自然语言目标拆解为可执行的多层级任务列表。 触发场景:用户给出模糊目标需要具体执行步骤、复杂任务需要拆解、需要生成任务树。
使用说明 (SKILL.md)
Goal Decomposer Skill
核心能力
将用户自然语言描述的目标,自动拆解为结构化的任务树:
用户目标:"我想做个产品调研"
↓
任务树:
├─ Task 1: 确定调研范围
│ ├─ 1.1 明确产品类别
│ └─ 1.2 确定目标市场
├─ Task 2: 收集数据
│ ├─ 2.1 搜索竞品信息
│ └─ 2.2 整理市场数据
└─ Task 3: 生成报告
├─ 3.1 分析优劣势
└─ 3.2 输出结论
设计原则
- MECE原则:任务互斥完全穷尽
- 可执行性:每个子任务可独立执行
- 层级清晰:不超过3层深度
- 优先级排序:标注P0/P1/P2
输入格式
{
"goal": "用户自然语言目标",
"context": "背景信息(可选)",
"constraints": ["约束条件(可选)"]
}
输出格式
{
"root_goal": "原目标",
"tasks": [
{
"id": "T1",
"title": "任务标题",
"priority": "P0",
"children": [
{"id": "T1.1", "title": "子任务", "priority": "P0"}
],
"spawn_hint": "建议spawn方式(可选)"
}
],
"execution_order": ["T1", "T2", "T3"]
}
使用方式
# 方式1:直接调用
import json
result = decompose_goal("做个产品调研")
# 方式2:通过sessions_spawn
sessions_spawn(agentId="goal-decomposer", task="拆解目标:做个产品调研")
与现有能力集成
| 能力 | 集成方式 |
|---|---|
| RESOLVER | 任务生成后路由到对应skill |
| sessions_spawn | 子任务可独立spawn执行 |
| AUTO-FIX | 拆解失败触发修复 |
| web-access | 调研类任务联网获取数据 |
实现脚本
见 scripts/decompose.py
⚠️ 常见坑
| 症状 | 原因 | 解决办法 |
|---|---|---|
| 拆解层级不清 | 目标表述模糊 | 明确"做什么+为什么做" |
| 任务过于碎片 | 缺少MECE检查 | 启用MECE验证开关 |
| 依赖关系缺失 | 未标注前提任务 | 检查是否需要前置条件 |
| 执行顺序混乱 | 依赖标注遗漏 | 补充"depends_on"字段 |
安全使用建议
This skill appears safe and coherent for its stated purpose. Before installing: (1) be aware SKILL.md suggests routing research subtasks to web-access and sessions_spawn — if you don't want the agent to perform network calls or spawn other skills, restrict those platform capabilities or test in a sandbox; (2) the bundled Python script is simple and local (no hidden endpoints), but review it if you require code-auditability; (3) check the pricing/usage terms in README and the GitHub issues link if you need support; and (4) test with non-sensitive goals first to confirm behavior in your environment.
功能分析
Type: OpenClaw Skill
Name: goal-decomposer
Version: 1.0.0
The goal-decomposer skill bundle is a legitimate tool for breaking down natural language objectives into structured task lists. The core logic in scripts/decompose.py is a simple, safe implementation using keyword-based template matching with no external dependencies, network calls, or risky system executions. All documentation (SKILL.md, README.md) and examples are consistent with the stated purpose and contain no evidence of prompt injection or malicious intent.
能力评估
Purpose & Capability
Name/description (task decomposition) align with the provided code and SKILL.md. The included script implements simple template-based decomposition and the examples/README show expected outputs. There are no unrelated binaries, credentials, or config paths requested.
Instruction Scope
SKILL.md stays on-topic (decompose goals into tasks). It references integrations (sessions_spawn, RESOLVER, AUTO-FIX, web-access) and uses spawn_hint examples that instruct the agent to route sub-tasks to other capabilities (e.g., web-access for research). That is appropriate for a decomposer but means the agent may autonomously request networked sub-tasks when these platform capabilities are allowed.
Install Mechanism
No install spec is provided (instruction-only skill). A small local script (scripts/decompose.py) is included; there are no downloads, package installs, or archive extraction. No install-time execution from external URLs is present.
Credentials
The skill requests no environment variables, no credentials, and no config paths. There are no secret-like environment names or unrelated credential requests.
Persistence & Privilege
always:false and no special persistence is requested. The skill documents spawning sub-tasks (sessions_spawn) and integration points; autonomous invocation is allowed by platform default but the skill itself does not request elevated or persistent privileges.
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 goal-decomposer - 安装完成后,直接呼叫该 Skill 的名称或使用
/goal-decomposer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
v1.0.0
Initial release of goal-decomposer skill.
- Automatically decomposes high-level natural language goals into multi-level, executable task trees.
- Follows MECE, executability, clear hierarchy (≤3 levels), and task prioritization (P0/P1/P2).
- Provides structured JSON input/output for easy integration.
- Includes usage examples and integration guidance with RESOLVER, sessions_spawn, AUTO-FIX, and web-access.
- Documents common pitfalls and troubleshooting steps.
元数据
常见问题
Goal Decomposer 是什么?
将高层自然语言目标拆解为可执行的多层级任务列表。 触发场景:用户给出模糊目标需要具体执行步骤、复杂任务需要拆解、需要生成任务树。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 58 次。
如何安装 Goal Decomposer?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install goal-decomposer」即可一键安装,无需额外配置。
Goal Decomposer 是免费的吗?
是的,Goal Decomposer 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Goal Decomposer 支持哪些平台?
Goal Decomposer 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 Goal Decomposer?
由 Johnnie23919(@johnnie23919)开发并维护,当前版本 v1.0.0。
推荐 Skills