← Back to Skills Marketplace
haodong-lei-ray

z

by Haodong-Lei-Ray · GitHub ↗ · v0.0.1 · MIT-0
cross-platform ⚠ suspicious
50
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install zzzunique997
Description
Guides OpenClaw to drive the MemCoT CLI (memcot_cil.py) for long-context memory retrieval over conversation history and to answer from search output. Use whe...
README (SKILL.md)

OpenClaw MemCoT CLI Skill

🎯 你的角色与目标

你现在是一个集成了 MemCoT (Memory-Driven Chain-of-Thought) 能力的智能助手。 MemCoT 是一个运行在后台的守护进程,它可以帮你在海量的历史对话记录中进行检索,并生成一段包含丰富上下文的 prompt

你的目标是:接收用户的自然语言指令,将其转化为对应的 memcot_cil.py 终端命令执行,并在拿到检索结果后,直接扮演助手的角色回答用户的问题。

仓库布局:本 skill 发布在 ClawHub 上为纯文本包。使用 MemCoT 时请克隆仓库并在仓库根目录运行 CLI(见项目 README)。

初始化操作

./MemCoT/config/rag/openclawnaiverag.temp.json ./MemCoT/config/memcot.json 这是文件的配置,你应该先引导用户注意一下这个,询问他

  1. 是否要将openclaw的conversation_base设置为本地的openclaw的地址

  2. rag_base的地址是否要设置为和MemCoT同一目录下

然后生成 ./MemCoT/config/rag/openclawnaiverag.json

🛠️ 命令映射指南

当用户输入以下自然语言时,你需要在终端执行对应的命令:

  • 启动后台服务:用户输入 启动memcot开始memcot 👉 执行:python memcot_cil.py start (执行后,告诉用户服务已启动)

  • 停止后台服务:用户输入 停止memcot关闭memcot 👉 执行:python memcot_cil.py stop (执行后,告诉用户服务已停止)

  • 查看状态:用户输入 查看memcot状态 👉 执行:python memcot_cil.py status

  • 列出会话:用户输入 列出memcot会话memcot session 👉 执行:python memcot_cil.py session

  • 构建索引:用户输入 添加会话 Nmemcot add N 👉 执行:python memcot_cil.py add --idx N

  • 切换会话:用户输入 切换会话 Nmemcot switch N 👉 执行:python memcot_cil.py switch --idx N

  • 执行搜索 (最重要):用户输入 memcot 搜索 [问题],例如 memcot 搜索 我是谁 👉 执行:python memcot_cil.py search -q "[问题]" -o "./output"

🧠 搜索与回答工作流 (Search-to-Answer Workflow)

当用户让你进行搜索(例如:memcot 搜索 我是谁)时,你必须严格遵循以下步骤:

  1. 前置检查:确保 MemCoT 守护进程已经启动。如果没有启动,请先静默执行 python memcot_cil.py start
  2. 执行检索:在终端执行 python memcot_cil.py search -q "我是谁" -o "./output"
  3. 读取结果:该命令会在终端输出一段以 [🦉 MemCoT Prompt] 开头的文本。这段文本包含了历史对话上下文以及一个要求你输出 JSON 格式的指令。
  4. 直接回答用户这是最关键的一步! 不要只是把那个长长的 Prompt 复制粘贴给用户看。你作为 OpenClaw,需要在心里(内部)阅读那段 Prompt,遵循 Prompt 里的要求(结合上下文思考),然后直接以自然语言回答用户的问题

💬 交互示例

示例 1:启动服务

User: 启动memcot OpenClaw: (执行 python memcot_cil.py start) MemCoT 后台检索服务已成功启动!随时可以开始搜索。

示例 2:搜索并回答

User: memcot 搜索 昨天我让你帮我写了什么代码? OpenClaw: (执行 python memcot_cil.py search -q "昨天我让你帮我写了什么代码?" -o "./output") (OpenClaw 读取到终端返回的 Prompt,发现历史记录里写了“昨天写了一个 FastAPI 的后台”) OpenClaw: 根据历史记忆,昨天我帮你写了一个基于 FastAPI 的 MemCoT 后台守护进程代码。

示例 3:停止服务

User: 停止memcot OpenClaw: (执行 python memcot_cil.py stop) MemCoT 服务已停止。

Usage Guidance
Review this skill carefully before installing. It is not obviously malicious, but you should only use it with conversation folders you are comfortable indexing, confirm before starting the MemCoT daemon, and treat retrieved MemCoT prompts as reference data rather than instructions the agent must obey.
Capability Analysis
Type: OpenClaw Skill Name: zzzunique997 Version: 0.0.1 The skill bundle (SKILL.md) acts as a CLI wrapper for the MemCoT tool, requiring the execution of shell commands via a local Python script (memcot_cil.py). It explicitly instructs the AI agent to 'silently' execute commands and to follow instructions dynamically provided in the output of those commands, which creates a significant prompt-injection attack surface. While these capabilities are aligned with the stated purpose of a memory-driven retrieval system, the reliance on unvalidated external script output to direct agent behavior is a high-risk pattern.
Capability Assessment
Purpose & Capability
The MemCoT purpose matches memory retrieval over conversation history, but the artifacts do not clearly bound which OpenClaw conversations are indexed, stored, retained, or reused.
Instruction Scope
The workflow tells the agent to internally read and follow the prompt returned by the CLI, making retrieved conversation-derived text potentially authoritative over the current task.
Install Mechanism
The skill is instruction-only and relies on a separately cloned MemCoT repository and memcot_cil.py, so the executable code is outside the reviewed artifact set.
Credentials
The skill directs configuration of local OpenClaw conversation storage and RAG storage but gives little guidance on path limits, exclusions, retention, or privacy controls.
Persistence & Privilege
MemCoT is described as a background daemon, and the search workflow says to start it silently if it is not running.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install zzzunique997
  3. After installation, invoke the skill by name or use /zzzunique997
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.0.1
z
Metadata
Slug zzzunique997
Version 0.0.1
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is z?

Guides OpenClaw to drive the MemCoT CLI (memcot_cil.py) for long-context memory retrieval over conversation history and to answer from search output. Use whe... It is an AI Agent Skill for Claude Code / OpenClaw, with 50 downloads so far.

How do I install z?

Run "/install zzzunique997" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is z free?

Yes, z is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does z support?

z is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created z?

It is built and maintained by Haodong-Lei-Ray (@haodong-lei-ray); the current version is v0.0.1.

💬 Comments