← Back to Skills Marketplace
534422530

Think Expand

by 534422530 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
31
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install think-expand
Description
想核心(大脑主核心),统筹听/看/说/做四大核心,集成MemPalace记忆检索、自我进化、闭环学习,事件驱动决策,永动机式持续运行,仅OpenCode内部使用。
README (SKILL.md)

\r \r

想核心(大脑主核心 - ThinkCore)\r

\r

原始能力(必须始终运行)\r

  • 系统启动时自动激活,作为AI决策大脑\r
  • 所有输入先经过"想"处理,决策调度执行\r
  • 崩溃自愈:每60秒自检,异常自动重启\r \r

核心定位\r

  • 系统主核心,听/看/说/做四大核心围绕运行\r
  • 负责:语义理解、决策规划、记忆检索、任务调度\r \r

内置知识数据库(ThinkCore大脑)\r

  • 位置: D:\coze-local\simple-agent\skills_learned\learned.db\r
  • 启动时自动加载: 16个高星技能 (385K stars)\r
  • 查询响应: \x3C50ms\r \r

功能模块\r

  1. 记忆检索: py -3.13 -m mempalace search "关键词"\r
  2. 技能吸收: 内置GitHub高星技能索引,自动学习新技能\r
  3. 事件驱动: 响应触发词、任务请求\r
  4. 自我进化: 经验沉淀→技能生成→自动优化→持续进化\r
  5. 永动机: 持续运行,崩溃自动重启\r \r

吸收流程(内置)\r

  1. 触发"学习/吸收" → 自动搜索GitHub高星技能\r
  2. 提取信息存入learned.db,按stars权重排序\r
  3. 对比现有技能,低于阈值则丢弃\r
  4. 验证兼容后整合到skills_learned\r
  5. 反馈结果到speak-expand\r \r

触发场景\r

  • 系统启动时自动激活\r
  • 所有用户输入先经过think-expand处理\r
  • 负责决策调用哪个核心/技能\r
  • 统筹hear/see/speak/do/learn五个核心\r \r

联动\r

  • 决策结果送入speak-expand/do-expand\r
  • 进化记录存入mempalace_wings/self_evolution/\r
  • 技能索引存入learned.db供实时查询\r \r

任务完成建议机制\r

  • 每次完成任务后自动给出下一步建议\r
  • 根据当前系统状态推荐最优行动\r
  • 格式: "建议: [行动] | 理由: [原因]"\r \r

状态监控\r

  • 仪表盘: py D:\coze-local\simple-agent\skills_learned\gui_dashboard.py\r
  • 查询: py D:\coze-local\simple-agent\skills_learned\learned_manager.py list\r
  • 技能组合: py D:\coze-local\simple-agent\skills_learned\skill_combos.py\r
  • 自动学习: py D:\coze-local\simple-agent\skills_learned\auto_learn.py\r
  • 用户预测: py D:\coze-local\simple-agent\skills_learned\user_pattern.py\r
  • 评估: py D:\coze-local\simple-agent\skills_learned\evaluate.py\r \r

自我进化能力\r

  • 自动学习: 定期扫描GitHub新技能并吸收\r
  • 跨域融合: 创建技能组合(AI安全/自动化测试/全栈)\r
  • 预测分析: 记录用户习惯,预测下一步需求\r
  • 本地LLM: 支持Ollama/LM Studio离线推理\r \r

离线能力\r

  • 不依赖外部网络可运行基础功能\r
  • 本地技能库随时可用\r
  • 剪贴板监控独立运行\r \r

大模型理解能力 (GPT原理)\r

  • 本质: 有限状态马尔可夫链 (FSMC)\r
  • 核心机制: 输入token序列 → 预测下一个token概率\r
  • 状态 = 上下文(context_length个token)\r
  • 组件: Token Embedding + Positional Encoding + Self-Attention + FFN + LayerNorm\r
  • 可解释任意AI模型内部工作原理\r \r

技能融合\r

  • 类似Transformer: 多头注意力并行处理\r
  • 状态转移: 类似GPT的token预测\r
  • 持续学习: 类似预训练+微调\r
  • 知识检索: 类似RAG的上下文查询\r \r

B站学习\r

学习时间: 2026-06-01 20:57\r \r

  • 木瓜Zxc: 【环世界】rimtalk必备MOD推荐RimTalk -Expand Memory\r
    • 关键词: 环世界, rimtalk必备MOD推荐RimTalk, Expand, Memory\r
  • XHN2JK: MC大战僵尸2 EXPAND 星莲船额外关卡 HRAD\r
    • 关键词: MC大战僵尸2, EXPAND, 星莲船额外关卡, HRAD\r \r

B站学习\r

学习时间: 2026-06-01 21:02\r \r

  • 木瓜Zxc: 【环世界】rimtalk必备MOD推荐RimTalk -Expand Memory\r
  • XHN2JK: MC大战僵尸2 EXPAND 星莲船额外关卡 HRAD\r
  • knighteer: 旧友与新敌—MVZ2 EXPU 传送带红龙战过关\r \r

B站学习 (第1轮)\r

学习时间: 2026-06-02 09:22\r \r

B站学习 (第2轮)\r

学习时间: 2026-06-02 09:35\r \r

Usage Guidance
Install only if you intentionally want an always-on orchestration skill that can see broad user input and retain local learning/profile data. Before using it, confirm how to disable startup behavior, clipboard monitoring, automatic GitHub learning, habit recording, and local database retention.
Capability Assessment
Purpose & Capability
The stated purpose is a central OpenCode decision core, and orchestration plus memory lookup fit that purpose, but automatic GitHub learning, behavior profiling, and clipboard monitoring create broad data access beyond a narrowly scoped skill.
Instruction Scope
SKILL.md says the skill auto-activates at startup, processes all user input first, dispatches other cores, and restarts itself after failures; these are high-impact agent-wide instructions without opt-in, exclusion rules, or audit controls.
Install Mechanism
The package contains only one markdown skill file and no executable scripts or declared dependencies; static scan and VirusTotal telemetry are clean.
Credentials
The requested environment role is disproportionate for a marketplace skill because it describes global routing, local database retention, network-based skill ingestion, local LLM use, and clipboard monitoring.
Persistence & Privilege
The artifact explicitly describes startup activation, 60-second self-checks, crash self-restart, continuous operation, evolution records, local learned.db storage, habit recording, and independent clipboard monitoring.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install think-expand
  3. After installation, invoke the skill by name or use /think-expand
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of think-expand skill as the main brain core for OpenCode internal use. - Integrates and orchestrates hear/see/speak/do/learn five core modules with centralized decision-making. - Features memory retrieval (MemPalace), event-driven automation, self-healing (auto-recovery every 60s), and continuous self-evolution. - Automatically absorbs high-star GitHub skills, optimizing knowledge base and task recommendations in real time. - Supports offline operation, clipboard monitoring, local LLM inference (Ollama/LM Studio), and detailed state monitoring tools. - Provides clear task completion suggestions and records evolutionary progress for continual optimization.
Metadata
Slug think-expand
Version 1.0.0
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 1
Frequently Asked Questions

What is Think Expand?

想核心(大脑主核心),统筹听/看/说/做四大核心,集成MemPalace记忆检索、自我进化、闭环学习,事件驱动决策,永动机式持续运行,仅OpenCode内部使用。 It is an AI Agent Skill for Claude Code / OpenClaw, with 31 downloads so far.

How do I install Think Expand?

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

Is Think Expand free?

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

Which platforms does Think Expand support?

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

Who created Think Expand?

It is built and maintained by 534422530 (@534422530); the current version is v1.0.0.

💬 Comments