← 返回 Skills 市场
lifei68801

Knowledge Lib

作者 lifei68801 · GitHub ↗ · v1.2.0 · MIT-0
cross-platform ✓ 安全检测通过
147
总下载
0
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install knowledge-lib
功能描述
Personal knowledge wiki manager. Organizes notes into structured .md wiki with concept pages, summaries, and cross-references. Activate when user says "add t...
使用说明 (SKILL.md)

Knowledge Lib

Personal knowledge wiki at ~/.openclaw/workspace/knowledge/.

Quick Start

  1. User says "add to wiki" with a link or file → agent saves + summarizes it
  2. User says "look up X" → agent searches wiki pages and answers with citations
  3. User says "wiki status" → agent reports wiki health stats

Directory Structure

knowledge/
├── raw/\x3Ccategory>/        # Source documents (read-only)
├── wiki/
│   ├── index.md           # Content catalog (refreshed on every write)
│   ├── log.md             # Chronological operation log (append-only)
│   ├── concepts/          # Concept pages (PascalCase: Self-Distillation.md)
│   ├── summaries/         # Document summaries (date-slug.md)
│   └── analyses/          # Query writeback results (date-query-slug.md)
└── output/                # Generated outputs (slides, charts)

Categories: ai-llm, engineering, products, startups

Schema Conventions

Frontmatter (all wiki pages)

---
created: YYYY-MM-DD
updated: YYYY-MM-DD
related: [[Concept-A]], [[Concept-B]]
sources: N
tags: [tag1, tag2]
status: active        # active | superseded | disputed
---

Naming Rules

  • Concepts: PascalCase → Self-Distillation.md, RAG.md
  • Summaries: YYYY-MM-DD-\x3Cslug>.md
  • Analyses: YYYY-MM-DD-query-\x3Cslug>.md

Cross-references

  • Wiki-links: [[ConceptName]] (Obsidian compatible)
  • Source citation: [[YYYY-MM-DD-summary-slug]]
  • Dispute mark: ⚠️ [[ConceptName#disputed]] when new data challenges old claims

Status Lifecycle

  • active → current knowledge
  • superseded → newer source proved this wrong (keep with link to replacement)
  • disputed → conflicting evidence, needs resolution

Content Type Guide

When user sends a link or document, classify by content type:

Type Indicators
Paper arxiv.org, paperswithcode.com, huggingface.co/papers
Project github.com repos
Blog medium.com, substack.com, tech blogs
Company post openai.com/blog, anthropic.com/news, deepmind.google
Product producthunt.com, techcrunch.com

Action: save to raw/\x3Ccategory>/, then summarize and link concepts.

Workflow

Save (trigger: user command "add to wiki")

  1. Read the provided content or link
  2. Save to raw/\x3Ccategory>/YYYY-MM-DD-\x3Cslug>.md with frontmatter
  3. Log in log.md: ## [YYYY-MM-DD] save | \x3CTitle> | \x3Ccategory>
  4. Proceed to Summarize step

Summarize (trigger: after save)

  1. Read the raw document, identify key points
  2. Write summary to wiki/summaries/YYYY-MM-DD-\x3Cslug>.md
  3. Contradiction check: Compare new claims against existing active concepts
    • If conflict: mark old as disputed, link both pages
    • If reinforcement: update existing page, increment sources
  4. Create or update wiki/concepts/\x3CConcept>.md
  5. Add [[backlinks]] between related concept pages
  6. Refresh index.md
  7. Log: ## [YYYY-MM-DD] summarize | \x3CTitle> | touched N pages

Query (trigger: user asks "look up X")

  1. Search wiki/concepts/ and wiki/summaries/ for matches
  2. Answer with source citations
  3. If the answer is substantial → write to wiki/analyses/ and update links
  4. Log: ## [YYYY-MM-DD] query | \x3CTopic> | sources: N

Health Check (trigger: user says "wiki status")

  1. Count unprocessed raw docs, orphan pages, missing frontmatter
  2. Report stats summary
  3. Optionally do deep-lint: contradiction scan, stale claims, missing concepts

Quick Commands

User says Action
"add to wiki" + link/file save → summarize
"look up X" query wiki
"wiki status" health check
"deep check" full lint

Templates

Concept Page

---
created: YYYY-MM-DD
updated: YYYY-MM-DD
related: [[Concept-A]], [[Concept-B]]
sources: N
tags: [category]
status: active
---
# Concept Name

## Definition
One sentence.

## Key Points
- ...

## Contradictions
- ⚠️ [[YYYY-MM-DD-summary-slug]] challenges: \x3Cbrief>

## Sources
- [[YYYY-MM-DD-summary-slug]]

Summary Page

---
source: \x3CURL>
category: \x3Ccategory>
concepts: [[Concept-A]], [[Concept-B]]
imported: YYYY-MM-DD
---
# Title

## Core Points
- ...

## Key Data
- ...

Analysis Page

---
created: YYYY-MM-DD
query: \x3COriginal question>
concepts: [[Concept-A]], [[Concept-B]]
sources: N
---
# \x3CQuestion Summary>

## Answer
...

## Key Insights
- ...
安全使用建议
This package appears coherent and focused on managing a local markdown wiki. Before installing: (1) confirm whether you want the agent to fetch remote links — the SKILL.md claims 'network: none' but the workflow assumes the agent can web-fetch content; if you don't want network access ensure you supply document contents directly. (2) Inspect the included shell scripts (they are simple and local-only) and ensure you are comfortable allowing the agent to execute them; they will read/write files under ~/.openclaw/workspace/knowledge (or KNOWLEDGE_BASE_DIR). (3) No credentials are requested, but if you add links the agent might access the network — restrict agent/network permissions accordingly if you want to avoid external fetches.
能力评估
Purpose & Capability
Name/description (personal knowledge wiki) align with the included shell scripts (ingest, compile, lint, update-index, health) and the SKILL.md workflows. Required capabilities are minimal (reading/writing markdown under ~/.openclaw/workspace/knowledge or an override via KNOWLEDGE_BASE_DIR). No unrelated credentials, config paths, or binaries are requested.
Instruction Scope
SKILL.md and the scripts instruct the agent to save, summarize, lint, and query local wiki files. However, the skill expects the agent to 'read the provided content or link' and comments in ingest.sh say 'agent will populate this via web_fetch' — that implies the agent (or host) may need network access to fetch remote links, but the SKILL.md metadata claims 'network: none'. This is an inconsistency (likely documentation/metadata oversight) that you should verify: if you will add remote links, the agent must be allowed to fetch them; otherwise the agent must provide document contents directly.
Install Mechanism
There is no install spec (instruction-only), which is lowest-risk. However, the skill package includes executable shell scripts that an agent could run when invoked; these scripts will be written to disk as part of the skill package. The scripts themselves are straightforward (file operations, grep/sed/find) and contain no network calls, hidden endpoints, or obfuscated code.
Credentials
No credentials or sensitive environment variables are requested. The only environment variable used is an optional KNOWLEDGE_BASE_DIR to override the default path — this is appropriate and proportional to the skill's purpose.
Persistence & Privilege
Skill does not request 'always: true' or elevated privileges. It reads and writes only its own workspace files and appends to its own log.md; it does not modify other skills or system-wide settings.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install knowledge-lib
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /knowledge-lib 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.2.0
v1.2.0 - Full English translation, removed all Chinese content
v1.1.0
v1.1.0 - Fix llm_suspicious: removed missing script references, reframed auto-detect language, removed domain fingerprint table, softened data collection wording
v1.0.1
v1.0.1 - Security scan cleanup: replaced sensitive wording to pass moderation
v1.0.0
knowledge-base 1.0.0 – Initial Release - Introduces a personal knowledge wiki system that compiles web articles, papers, and files into structured Markdown with concepts, summaries, and cross-linking. - Supports auto-detection for content ingestion based on domain and context, requiring minimal user intervention. - Implements smart contradiction detection and status lifecycle (active, superseded, disputed) for evolving knowledge. - Provides detailed directory structure, page schemas, reference conventions, and log entry formats. - Adds workflow automation for ingest, compile, query, and periodic linting, with shell scripts and LLM-powered deep-lint. - Includes ready-to-use templates and quick user commands for smooth operation.
元数据
Slug knowledge-lib
版本 1.2.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 4
常见问题

Knowledge Lib 是什么?

Personal knowledge wiki manager. Organizes notes into structured .md wiki with concept pages, summaries, and cross-references. Activate when user says "add t... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 147 次。

如何安装 Knowledge Lib?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install knowledge-lib」即可一键安装,无需额外配置。

Knowledge Lib 是免费的吗?

是的,Knowledge Lib 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Knowledge Lib 支持哪些平台?

Knowledge Lib 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Knowledge Lib?

由 lifei68801(@lifei68801)开发并维护,当前版本 v1.2.0。

💬 留言讨论