← 返回 Skills 市场
Cognitive Topology
作者
qiushang555
· GitHub ↗
· v1.0.0
· MIT-0
117
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install cognitive-topology
功能描述
Agent认知拓扑引擎。自动将复杂任务分裂为多个独立分支,并行执行后汇总整合。触发场景:(1)复杂问题需要多角度分析 (2)多独立子任务需要并行执行 (3)需要清晰追踪思维分支 (4)用户说"分叉"、"分支分析"、"多线程"时。使用 sessions_spawn 创建分支,分支结果写入 L2 文件,Main Se...
使用说明 (SKILL.md)
Cognitive Topology
把复杂任务的「一条直线」变成「多分支并行」。
何时触发
当任务符合以下任一条件时,自动启用本技能:
- 需要同时分析 N 个不同方向/角度
- 任务可拆分为相互独立的子任务
- 问题复杂,需要递归分解
- 用户明确说「分叉」「分支」「多角度」
核心工作流
Step 1: Fork - 创建分支
使用 sessions_spawn 创建独立分支:
sessions_spawn(
mode="session",
runtime="subagent",
task="\x3C具体分支任务>",
label="ct-\x3C序号>-\x3C简短描述>"
)
每个分支有独立 Session ID,互不干扰。
Step 2: Branch - 分支执行
- 分支 Agent 独立执行自己的任务
- 完成后将结论写入标准 L2 文件:
~/.openclaw/workspace/cognitive-topology/branches/\x3Cbranch_id>_L2.md - L2 格式:## 结论
\x3C具体结论>
\
依据\
\x3C分析依据>
Step 3: Integrate - 整合
Main Session 等待所有分支完成,读取所有 L2 文件,生成 synthesis:
cat ~/.openclaw/workspace/cognitive-topology/branches/*_L2.md
目录结构
~/.openclaw/workspace/cognitive-topology/
├── branches/ # 分支工作目录
│ ├── ct-1_anal_ai_L2.md # 分支1的L2
│ ├── ct-2_anal_chip_L2.md # 分支2的L2
│ └── ...
├── topology_latest.json # 当前拓扑状态
└── synthesis.md # 整合结果
关键约束
- 分支只写 L2,不读自己的 L2
- Main Session 读所有 L2,不直接读分支的 Session 历史
- 分支间不直接通信
- 跨分支信息通过 Main Session 的 synthesis 传播
简化实现说明
无 UI,星空图用文本树替代:
🌐 认知拓扑
├── branch-1 [分析AI板块]
│ └── ✅ 已完成
├── branch-2 [分析半导体]
│ └── ⏳ 进行中
└── branch-3 [分析新能源]
└── ⏳ 待启动
典型场景示例
场景:分析 A 股三个方向的机会
Main Session 收到任务后:
- 判断需要分裂为3个独立分支
- 同时 spawn 3 个分支 Agent
- 分支1分析AI板块 → 写 L2
- 分支2分析半导体 → 写 L2
- 分支3分析新能源 → 写 L2
- Main 汇总 L2 → 生成 synthesis → 回复用户
安全使用建议
This skill appears to do what it says, but review and possibly modify file paths and permissions before enabling: 1) Confirm the runtime user and adjust hard-coded '/root/.openclaw/...' paths to the correct non-root workspace (or accept that root is required). 2) Be aware that branch outputs and syntheses are appended to MEMORY.md and daily snapshot files — do not allow sensitive secrets in branch outputs. 3) Test in an isolated environment first to verify behavior (especially ct_archive which reads/appends to MEMORY.md and may assume the file exists). 4) If you don't want permanent storage, change the scripts to use a temporary workspace or add retention/cleanup logic. 5) If you rely on sessions_spawn, ensure your agent's spawning policy/privileges are appropriate.
功能分析
Type: OpenClaw Skill
Name: cognitive-topology
Version: 1.0.0
The 'cognitive-topology' skill is a task orchestration framework designed to decompose complex tasks into parallel sub-tasks (branches) using the OpenClaw 'sessions_spawn' capability. The included Python scripts (ct_fork.py, ct_integrate.py, ct_archive.py, etc.) manage the lifecycle of these branches by creating, tracking, and aggregating results into a synthesis file and a long-term memory log within the local workspace directory (/root/.openclaw/workspace/). The logic is transparent, lacks network activity or obfuscation, and operates strictly within the expected agent environment boundaries.
能力评估
Purpose & Capability
The name/description (split tasks into branches, write L2 files, integrate results) matches the included scripts (ct_fork, ct_complete, ct_integrate, ct_archive, ct_status). There are no unrelated credentials or external services requested. One implementation detail: scripts use hard-coded /root/.openclaw paths while SKILL.md sometimes uses ~/.openclaw — this is functionally consistent but brittle and privileges/path-sensitive.
Instruction Scope
SKILL.md instructs the agent to spawn sessions and read/write L2 files under the workspace, which aligns with the scripts. Concerns: (1) path inconsistency between '~/.openclaw' in the documentation and '/root/.openclaw' in scripts may cause incorrect file locations or require root privileges; (2) the skill instructs agents to run local scripts and to cat workspace files — it does not request unrelated system files or network exfiltration, but it does direct writing and reading of persistent files in the agent workspace.
Install Mechanism
Instruction-only install with included Python scripts — no external downloads, package managers, or archives. Low install risk because nothing is fetched from arbitrary URLs.
Credentials
No environment variables or credentials are requested. Scripts operate on local filesystem only. Note: writing to /root/.openclaw/workspace and /root/.openclaw/workspace/MEMORY.md persists outputs — consider whether those files may include sensitive user data.
Persistence & Privilege
always:false and no cross-skill config writes. The skill creates/updates persistent files (topology_latest.json, branch L2 files, synthesis.md, MEMORY.md and daily snapshots) under the workspace; this is expected for a memory/aggregation tool but does mean outputs are stored permanently on disk and may accumulate sensitive content.
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 cognitive-topology - 安装完成后,直接呼叫该 Skill 的名称或使用
/cognitive-topology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: Agent认知拓扑引擎,支持多分支并行执行和长期记忆整合
元数据
常见问题
Cognitive Topology 是什么?
Agent认知拓扑引擎。自动将复杂任务分裂为多个独立分支,并行执行后汇总整合。触发场景:(1)复杂问题需要多角度分析 (2)多独立子任务需要并行执行 (3)需要清晰追踪思维分支 (4)用户说"分叉"、"分支分析"、"多线程"时。使用 sessions_spawn 创建分支,分支结果写入 L2 文件,Main Se... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 117 次。
如何安装 Cognitive Topology?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install cognitive-topology」即可一键安装,无需额外配置。
Cognitive Topology 是免费的吗?
是的,Cognitive Topology 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Cognitive Topology 支持哪些平台?
Cognitive Topology 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 Cognitive Topology?
由 qiushang555(@qiushang555)开发并维护,当前版本 v1.0.0。
推荐 Skills