御知库

御知库——个人知识库系统。当帝提问时，优先从知识库检索答案。支持文档存储（Markdown/PDF/TXT）、语义搜索、关键词搜索、自动分类。帝说"查一下xxx"、"/知识库 xxx"或询问知识类问题时触发。

Security Scan

The skill is internally consistent with a local personal knowledge-base: it only reads files from a small set of user directories, stores data under ~/.yuzhi, and provides local search/CLI functionality with no network calls or secret requests.

Purpose & Capability

Name/description describe a personal KB. The included CLI and Python code implement adding, indexing, and searching local documents and only requests filesystem access consistent with that purpose.

Instruction Scope

SKILL.md and scripts describe only adding/listing/searching docs under ~/.yuzhi, Desktop, Documents, Downloads, and workspace dirs. Instructions do not reference unrelated system files, environment variables, or external endpoints.

Install Mechanism

There is no install spec and the skill is instruction + a small local script. No downloads, package installs, or external code retrieval are performed by the skill itself.

Credentials

The skill requires no environment variables or credentials. It creates/stores data under ~/.yuzhi and reads only from a limited set of user directories, which is proportional to a KB tool.

Persistence & Privilege

always is false and the skill does not request permanent elevated presence. It creates its own directory (~/.yuzhi) and an SQLite DB, which is expected for a local KB.

Assessment

This skill appears to do exactly what it says: a local personal knowledge base that stores data under ~/.yuzhi and reads files from Desktop/Documents/Downloads and certain workspace dirs. Before installing or using it, note: (1) it will create ~/.yuzhi and an SQLite index file (index.db); (2) PDF extraction runs a subprocess that requires the pdfplumber Python package—if pdfplumber is absent PDF text extraction will skip or fail; (3) only files in the allowed directories can be added, so sensitive files elsewhere won't be touched unless you move them into those locations; (4) avoid adding confidential documents unless you trust the environment, and inspect the script yourself if you have concerns. No network calls or secret access are requested by the skill.

御知库（yuzhi）

个人知识库系统，AI 优先检索，帝提问时自动召回。

核心架构

~/.yuzhi/
├── docs/              # Markdown 文件
├── uploads/           # PDF/TXT 等
├── metadata.jsonl     # 元数据（路径/标签/摘要/时间）
└── index.db          # SQLite 向量索引

# 搜索时 AI 优先查这里

CLI 命令

yuzhi add <文件路径> --tag <标签>

添加文件到知识库，自动提取摘要和标签。

yuzhi search <查询内容> --mode <semantic|keyword|all>

双轨搜索（语义+关键词），返回最相关的文档和内容片段。

yuzhi list --tag <标签> --limit <数量>

列出知识库文件，支持按标签筛选。

yuzhi stats

查看知识库统计：文件数、标签分布、存储大小。

工作流程

帝提问 → 触发检索 → yuzhi search → AI 优先用知识库内容回答

帝说"存一下这个" → yuzhi add → 自动分类打标签

标签体系

• 制度 — 朝廷制度、流程规范
• 项目 — 各项目文档、代码规范
• 决策 — 帝旨、重要决策记录
• 人物 — 人物资料、背景
• 技术 — 技术方案、架构文档
• 笔记 — 随手笔记、想法
• 其他 — 不属于以上的

检索优先级

1. 语义相似度（embedding cosine similarity）
2. 关键词命中（SQLite FTS5）
3. 时间权重（近期优先）