Taizi Knowledge Base

个人知识库 - 融合向量检索、实体关系、笔记管理

Security Scan

The SKILL.md claims a local personal knowledge-base but instructs running a specific Python script and touching user home directories without providing the script, required binaries, or clear configuration — the instructions and declared requirements are inconsistent.

Purpose & Capability

The skill describes a personal knowledge base (vector search, entities, notes) but declares no required binaries or files while the instructions explicitly call a Python script (vector_kb.py) at a hard-coded path. That script is not included and no Python requirement is declared, so the claimed capability cannot be realized as described.

Instruction Scope

Runtime instructions tell the agent to read and write files under user home paths (~/.openclaw and C:\Users\Administrator\...), add files to an Obsidian directory, and extract entities. Those actions involve accessing local filesystem data and modifying user notes; the instructions give broad filesystem targets and hard-coded, platform-specific paths (Windows Administrator) rather than a configurable vault path.

Install Mechanism

There is no install spec (instruction-only), which limits installer risk. However, the runtime depends on a local script that is not provided; the absence of an install step or included code means the skill will either fail at runtime or rely on pre-existing, opaque local scripts.

Credentials

The skill requests no credentials or environment variables (reasonable), but it also fails to declare the need for Python or any NLP/embedding service keys that entity extraction or vector indexing might require. The lack of declared dependencies is inconsistent with the described functionality.

Persistence & Privilege

always:false and no special OS restrictions are set. The skill does instruct filesystem writes at runtime, but it does not request permanent elevated platform privileges in its metadata.

What to consider before installing

This skill's instructions expect a local Python script (C:\Users\Administrator\.openclaw\scripts\vector_kb.py) and write/read locations under ~/.openclaw and an Obsidian vault, but the package contains no code and declares no dependencies. Before installing or using it: (1) ask the publisher for the actual script/source and an install plan; (2) do not run commands that write into your home or notes until you inspect the script contents; (3) confirm which OS/path the skill targets (it mixes Windows and Unix paths) and request configurable vault paths; (4) ensure Python and any required NLP/embedding services are explicitly declared; (5) if you must test, run in a sandboxed account or VM and review the script for data-exfiltration or unexpected behavior.

个人知识库

统一管理：向量检索 + 实体关系 + 笔记存储

触发词

• "存入知识库"
• "查知识库"
• "记一笔"
• "帮我记住"

存储位置

• 向量库：~/.openclaw/knowledge_base/
• 实体库：~/.openclaw/workspace/memory/ontology/
• 笔记库：~/.openclaw/workspace/notes/

命令

存入知识库

python C:\Users\Administrator\.openclaw\scripts\vector_kb.py add --text "内容" --source "来源"

搜索知识库

python C:\Users\Administrator\.openclaw\scripts\vector_kb.py search --query "关键词" --top 5

查看统计

python C:\Users\Administrator\.openclaw\scripts\vector_kb.py stats

添加笔记文件

python C:\Users\Administrator\.openclaw\scripts\vector_kb.py add-file --path "C:\path\to\file.md"

工作流程

1. 收到用户内容 → 判断类型
2. 实体关系 →存入 ontology
3. 文本内容 → 存入向量库
4. 笔记文件 → 存入 Obsidian 目录

自动处理

• 文本：自动分块存入向量库
• 实体：自动提取人物/项目/事件关系
• 文件：自动索引并可检索