Token导航 LogoToken导航

Agent Skill · 研究检索

lhuaizhong-llm-wiki(lhuaizhong LLM wiki 搜索)

lhuaizhong-llm-wiki 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

lhuaizhong-llm-wiki 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 26,214。

总安装 3,612周安装 152GitHub Stars 公开资料未说明下载 1,265核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
0.1.1

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:lhuaizhong-llm-wiki(lhuaizhong LLM wiki 搜索)
来源仓库:https://github.com/lhuaizhong/lhuaizhong-llm-wiki
安装命令:
openclaw skills install lhuaizhong-llm-wiki
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install lhuaizhong-llm-wiki

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

构建并维护本地 Markdown 知识维基,支持长期复合检索。适合在 OpenClaw 中根据关键词、任务场景或来源线索快速定位信息。通过关键词、仓库或 README 内容筛选候选结果,提升信息获取效率。

安装命令:openclaw skills install lhuaizhong-llm-wiki;建议确认权限与维护状态。注意可能涉及联网、文件读写或命令执行,需评估安全风险与资源占用。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install lhuaizhong-llm-wiki;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

lhuaizhong-llm-wiki 怎么安装?

在终端执行 openclaw skills install lhuaizhong-llm-wiki。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

lhuaizhong-llm-wiki 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

lhuaizhong-llm-wiki 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

lhuaizhong-llm-wiki 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

lhuaizhong-llm-wiki 用的人多吗?

公开数据是累计安装 3,612 次,近一周安装 152 次,下载 1,265 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

lhuaizhong-llm-wiki 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
lhuaizhong-llm-wiki

lhuaizhong LLM wiki 搜索 · 本页

3,612—需要联网MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
llm-wiki
description
Build and maintain a local Markdown knowledge wiki that compounds over time instead of relying on one-shot RAG retrieval. Use when the user wants to turn notes, documents, transcripts, chat logs, research material, or project context into an LLM-maintained wiki/second-brain; when setting up an Obsidian-friendly knowledge base; or when defining ingest, query, reindex, and lint workflows for a personal or team knowledge vault.

LLM Wiki

Overview

Build a local-first, Markdown-based wiki where the human curates sources and asks questions, while the agent maintains the compiled knowledge pages, links, summaries, and indexes.

Prefer this pattern when the goal is cumulative understanding across many sessions, files, and questions — not just answering one query from raw documents.

Core Idea

Treat the wiki as a compiled artifact, not the raw archive.

Use three layers:

  1. Raw sources: immutable inputs such as notes, PDFs, transcripts, copied text, logs, screenshots, and chat exports.
  2. Wiki pages: curated Markdown pages written and maintained by the agent.
  3. Schema/rules: lightweight conventions for page layout, naming, linking, and maintenance.

The human focuses on sourcing, exploration, and direction. The agent handles summarizing, cross-referencing, page updates, consolidation, and bookkeeping.

Recommended Directory Shape

Adapt the exact layout to the user’s workspace, but prefer a structure like:

knowledge/
  raw/
    inbox/
    docs/
    transcripts/
    web/
  wiki/
    concepts/
    people/
    projects/
    timelines/
    sources/
  logs/
    knowledge-log.md
  INDEX.md
  SCHEMA.md

Guidelines:

  • Keep raw/ append-only when practical.
  • Keep wiki/ human-readable and link-heavy.
  • Keep INDEX.md short and navigational.
  • Keep SCHEMA.md focused on conventions, not prose.

Operation Modes

1. Ingest

Use when new material arrives.

Goal: convert raw inputs into durable wiki structure.

Steps:

  1. Inspect new source material.
  2. Decide whether it belongs in raw/ unchanged.
  3. Extract stable entities: concepts, people, projects, decisions, methods, events, open questions.
  4. Create or update wiki pages.
  5. Add backlinks and related-page links.
  6. Update INDEX.md if navigation changed.
  7. Append a concise entry to the knowledge log.

Prefer updating existing pages over creating near-duplicates.

2. Query

Use when the user asks a question about the knowledge base.

Steps:

  1. Search the wiki first.
  2. Read the most relevant compiled pages.
  3. Fall back to raw sources only if the wiki is missing detail.
  4. Answer the question.
  5. If the answer revealed a gap, update the wiki before finishing when appropriate.

Prefer reading 2–5 good pages over sweeping the entire vault.

3. Reindex

Use when the structure has drifted or grown.

Tasks:

  • Refresh INDEX.md
  • Merge overlapping pages
  • Rename pages that no longer fit
  • Repair broken links
  • Normalize tags or page sections
  • Split oversized pages when they cover multiple topics

4. Lint

Use when the wiki may be internally inconsistent.

Check for:

  • duplicate pages for the same concept
  • contradictory claims
  • stale summaries after new evidence
  • orphan pages with no inbound or outbound links
  • pages that lack source pointers
  • vague claims that should be marked as uncertain or disputed

When high-stakes accuracy matters, preserve disagreement explicitly instead of forcing one synthetic summary.

Page Design Rules

Prefer simple Markdown that Obsidian and other editors render well.

For most concept pages, use sections like:

# Topic Name

## Summary
A short stable summary.

## Key points
- ...

## Relationships
- Related: [[Other Page]]
- Depends on: [[Another Page]]

## Sources
- [[Source - Interview 2026-04-06]]
- raw/transcripts/interview-2026-04-06.md

## Open questions
- ...

Rules:

  • Put the most reusable summary near the top.
  • Prefer explicit links like [[Page Name]] when the vault supports wiki links.
  • Separate facts, interpretations, and open questions when ambiguity matters.
  • Record provenance whenever practical.
  • Avoid decorative formatting and long narrative filler.

Indexing and Logging

Maintain two special files:

INDEX.md

Use as a map of the vault.

Include:

  • major sections
  • important hub pages
  • canonical names for key topics
  • pointers to active or recently changed areas

Do not turn it into a giant dump of every page.

logs/knowledge-log.md

Use as a chronological maintenance trail.

Each entry should briefly record:

  • what changed
  • why it changed
  • which pages were added or updated
  • any unresolved ambiguity

Example:

## 2026-04-06
- Ingested interview transcript on home networking setup.
- Updated [[OpenClaw 微信 channel 插件项目]] and created [[网络规划偏好]].
- Flagged uncertainty around deployment target and auth flow.

Workflow Heuristics

  • Prefer modifying a small number of relevant pages over regenerating the whole wiki.
  • Create new pages when a concept is likely to recur.
  • Merge pages when names differ but the concept is clearly the same.
  • Preserve raw material; do not destroy source context unless the user asks.
  • If the wiki disagrees with a source, update the wiki and mark the change.
  • If evidence is mixed, preserve the disagreement explicitly.
  • Use the wiki as the primary retrieval surface; use raw sources as ground truth.

Obsidian-Friendly Usage

If the user keeps notes in Obsidian:

  • write plain Markdown files
  • preserve wikilinks where possible
  • prefer stable filenames over clever ones
  • keep pages short enough to browse comfortably
  • use hub pages for dense topics instead of giant monoliths

Obsidian is the browsing interface; the agent is the maintainer.

Suggested User Requests That Should Trigger This Skill

Examples:

  • “把这些零散笔记整理成知识库”
  • “帮我做一个 second brain / wiki”
  • “把这些聊天记录沉淀成长期知识”
  • “给这个 Obsidian 仓库设计 ingest/query/lint 流程”
  • “把原始资料编译成可维护的 Markdown wiki”
  • “帮我重构这个知识库结构,减少重复页面”

Quick Start

For a brand-new vault, initialize the scaffold first:

python3 scripts/init_wiki.py /path/to/knowledge

Add --force to overwrite existing scaffold files.

After initialization:

  1. review SCHEMA.md
  2. place or copy source material into raw/
  3. ask the agent to ingest the material into wiki/
  4. use query / reindex / lint as the vault grows

References

Read these bundled references only when needed:

  • references/scaffold.md: example scaffold for SCHEMA.md, INDEX.md, and page conventions
  • references/operations.md: practical checklists for ingest, query, reindex, and lint runs
  • references/publish-notes.md: short external-facing positioning and use-case wording

Bundled Script

  • scripts/init_wiki.py: create a starter wiki vault with directories, hub pages, INDEX.md, SCHEMA.md, and logs/knowledge-log.md

平台分布

OpenClaw

87.72%
按下载量换算1,110

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills