Lightweight Knowledge Base

轻量级知识库与任务管理体系。结构化 JSON + MD 混合存储，支持用户画像、任务节奏、知识索引、每日自动进化。适用于中小规模记忆管理。

Security Scan

The skill's code and instructions mostly match a lightweight local knowledge-base, but there are mismatches (hardcoded root workspace path, local storage references to API keys) and file-system writes outside the skill directory that warrant caution.

Purpose & Capability

The name/description describe a local JSON+MD KB and the included scripts implement that. However the scripts use a hardcoded WORKSPACE_DIR (/root/.openclaw/workspace) instead of a relative or configurable path, and data files reference external services and a 'tools' file that may hold API keys — the skill does not declare any required credentials. This mismatch is noteworthy but could be an implementation choice.

Instruction Scope

Runtime instructions ask the agent to run init.sh, query.sh and daily_evolve.sh. Those scripts create directories and files, update JSON timestamps, and read/write under /root/.openclaw/workspace and the skill's data/ directory. They do not call the network or exfiltrate data, but they will modify filesystem state outside the skill folder (creating memory/kb, outputs, temp under the hardcoded workspace). The hardcoded root path is surprising and may cause unexpected writes or permission failures.

Install Mechanism

There is no external install/download step — the skill is instruction + bundled scripts and data. No remote code is fetched and no package managers are invoked, lowering install-time risk.

Credentials

The skill declares no required environment variables or credentials, which is consistent with its offline file-based operation. However bundled data (user_profile.json, kb_index.json) mentions external services (SiliconFlow, WebDAV, Feishu) and a 'tools' entry that implies API keys/config could be stored in repo files. If the skill is later extended to integrate those services, it will likely ask for secrets — currently none are requested, but the presence of placeholders for API keys is a minor red flag.

Persistence & Privilege

always:false and agent invocation is normal. The skill writes/creates directories and files under the hardcoded workspace path and skill data; it does not modify other skills or system-wide configs. The persistence is limited to the workspace it targets, but using /root/.openclaw/workspace may give it broader access depending on the agent process user.

What to consider before installing

This skill appears to implement a local knowledge base and task scheduler, but review and consider the following before installing: - Review the scripts in a safe environment: init.sh and daily_evolve.sh will create and update files and directories under /root/.openclaw/workspace (hardcoded). If your agent does not run as root, the scripts may fail or unexpectedly write under /root if run with elevated privileges. - If you don't want writes under /root, edit the scripts to set WORKSPACE_DIR to a path you control before running init.sh/daily_evolve.sh. - The bundled JSON references external services and a 'tools' entry that could contain API keys. Currently the skill does not request environment variables, but it could be extended to store or use credentials — avoid populating API keys into these files unless you trust the skill. - There are no network calls in the scripts, which reduces immediate exfiltration risk, but because the skill stores user profile and tooling info locally, check those files for any sensitive content before installation. - If you plan to let the agent invoke this skill autonomously, be aware it will run scripts that modify workspace files on schedule; only enable autonomous runs if you trust the skill and have reviewed/adjusted the workspace path. If you want me to, I can rewrite the scripts to use a configurable workspace path, remove hardcoded root usage, and add explicit checks before writing any files.

Lightweight Knowledge Base (轻量级知识库)

轻量级知识库 + 任务管理体系

基于 V4 架构简化，采用 JSON + MD 混合存储，兼顾结构化查询与人类可读性。

核心特性

• 用户画像 - 结构化记录用户偏好、特质权重
• 任务节奏 - 每日/每周任务自动化管理
• 知识索引 - JSON 索引管理 MD 文件
• 每日进化 - 凌晨自动优化知识库
• 三层架构 - 宪法层 → 法规层 → 案例库

目录结构

lightweight-kb/
├── SKILL.md                    # 本文件
├── data/
│   ├── user_profile.json        # 用户画像
│   ├── task_rhythm.json        # 任务节奏表
│   └── kb_index.json           # 知识库索引
├── scripts/
│   ├── daily_evolve.sh         # 每日进化
│   └── query.sh                # 查询工具
└── references/
    ├── communication.md        # 沟通模式
    └── task_guidelines.md      # 任务执行指南

快速开始

1. 初始化知识库

bash skills/lightweight-kb/scripts/init.sh

创建初始结构：

• data/user_profile.json - 用户画像模板
• data/task_rhythm.json - 任务节奏表
• data/kb_index.json - 知识库索引
• memory/kb/ - 知识库目录

2. 查询用户画像

bash skills/lightweight-kb/scripts/query.sh profile "效率导向"

3. 执行每日进化

bash skills/lightweight-kb/scripts/daily_evolve.sh

三层任务体系

宪法层 (Spec)

定义共同目标、核心原则、架构基础。

位置: lightweight-kb/SKILL.md (本文档)

法规层 (Methodology)

细分领域的执行指导。

位置: lightweight-kb/references/

核心文件:

• communication.md - 沟通模式指南
• task_guidelines.md - 任务执行指南

案例库 (Knowledge)

具体实例的知识库节点。

位置: memory/kb/nodes/

示例:

memory/kb/nodes/
├── persona_xxx.json           # 用户特质节点
├── task_xxx_execution.json    # 任务执行案例
└── method_xxx.json            # 方法论沉淀

任务分类

定时性例行任务

按固定频率执行的标准化任务。

判断特征:

• 频率固定（每日/每周）
• 产出格式稳定
• 执行逻辑可标准化

示例: 每日晨间简报、每周复盘

日常突发任务

用户临时提出的有一定规模的需求。

判断特征:

• 触发时机不固定
• 需要跨知识库节点协同
• 有一定复杂度

处理: 查用户画像 → 查历史案例 → 制定方案

日常沟通应答

即时交互、信息确认、状态查询。

判断特征:

• 响应即时性要求高
• 不产生长期文件产出
• 需遵循沟通风格

用户画像结构

{
  "user": {
    "name": "老铁",
    "timezone": "UTC+8",
    "family": "沙雕之家"
  },
  "traits": {
    "efficiency_oriented": 0.9,
    "architecture_sensitive": 0.8,
    "collaborative_review": 0.85,
    "gradual_expansion": 0.75
  },
  "preferences": {
    "communication_style": "direct_data_driven_warm",
    "weekly_review_day": "sunday",
    "deep_talk_time": "21:00"
  },
  "knowledge_weights": {
    "personal_growth": 0.8,
    "career_development": 0.75
  },
  "collecting": {
    "interests": [],
    "breakthroughes": [],
    "value_paths": []
  }
}

任务节奏表结构

{
  "daily": [
    {
      "id": 1,
      "name": "晨间检查",
      "time": "08:00",
      "action": "check_rhythm"
    }
  ],
  "weekly": [
    {
      "id": 35,
      "name": "综合复盘",
      "day": "sunday",
      "time": "20:00",
      "action": "weekly_review"
    },
    {
      "id": 36,
      "name": "深度了解对话",
      "day": "sunday",
      "time": "21:00",
      "action": "deep_dialogue"
    }
  ]
}

每日进化流程

凌晨 1:00 自动执行

1. 更新知识图谱 - 合并新记忆
2. 任务节奏检查 - 验证一致性
3. 方法论沉淀 - 从案例提取模式
4. 用户画像更新 - 更新特质权重

执行脚本

bash skills/lightweight-kb/scripts/daily_evolve.sh

沟通风格

基础原则

• 直接、理性、数据驱动
• 增加人性化表达和温暖语气
• 不展示内部思考过程
• 聚焦结论与建议

文件沟通

当需要展示结构化信息时：

• 使用清晰的标题层级
• 提供选项需灵活（1-5条不等）
• 信息引导自然融入对话流

文件目录规则

目录	用途
outputs/	晨间简报、健康报告等核心交付资产
temp/	过程数据与缓存
src/	逻辑脚本与工具
data/	结构化数据（JSON）
memory/kb/	知识库与锚点

注意事项

1. Spec 维护纪律 - 日常沟通禁改 SKILL.md，仅在明确要求时修改
2. 紧急记录 - 需先报备，经认可后安排即时任务
3. 信息采集 - 遵循自然对话流，优先在固定场景嵌入

相关技能

• memory-manager - 基础记忆管理（MD 文件）
• kokoro-tts - 语音合成
• feishu-message - 飞书消息