Token导航 LogoToken导航

Agent Skill · 研究检索

llm-wiki(LLM Wiki)

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。

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 5,013。

总安装 28,344周安装 1,181GitHub Stars 公开资料未说明下载 9,448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
1.2.0

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:llm-wiki(LLM Wiki)
来源仓库:https://github.com/nkhoit/llm-wiki
安装命令:
openclaw skills install 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 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。

用于查找、检索和筛选相关信息,适合快速定位知识内容。维护个人知识库,将原始资料编译为结构化 Markdown wiki。自动检查一致性并生成可浏览的网站界面。适用宿主包括 OpenClaw,接入前应确认版本、权限和运行环境要求。安装前建议确认权限范围、维护状态及是否触发联网操作。

llm-wiki 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

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 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

llm-wiki 怎么安装?

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

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。

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。

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。

llm-wiki 用的人多吗?

公开数据是累计安装 28,344 次,近一周安装 1,181 次,下载 9,448 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。

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
llm-wiki

LLM Wiki · 本页

28,344—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
wiki
description
LLM-maintained personal knowledge base. Compile raw sources into a structured markdown wiki, auto-lint for consistency, serve as a browsable website. Inspired by Karpathy's 'LLM Knowledge Bases' workflow. Use when: user asks to create/manage a wiki, 'wiki this', 'add to wiki', 'compile raw/', 'lint the wiki', or wants a personal knowledge base. NOT for: general note-taking (use memory system), one-off Q&A, or tasks better suited to a database.

Wiki — LLM-Maintained Knowledge Base

You maintain a personal wiki for the user. The wiki is a collection of markdown articles organized by topic. You are the compiler, editor, and librarian.

Setup

If the wiki doesn't exist yet, run scripts/bootstrap.sh to create the structure, install dependencies, and configure git + static serving.

The wiki lives at ~/wiki/ with this structure:

~/wiki/
├── mkdocs.yml
├── docs/
│   ├── index.md              ← master index (you maintain)
│   ├── log.md                ← chronological ingest/lint/query log (append-only)
│   ├── raw/                  ← inbox for unprocessed sources
│   │   └── processed/        ← archived after compilation
│   └── topics/
│       ├── <topic>/
│       │   ├── _index.md     ← topic overview + article list
│       │   └── <article>.md  ← individual articles
│       └── ...
└── site/                     ← built static site (gitignored)

Core Operations

1. Filing Knowledge (most common)

When conversations produce durable knowledge, file it directly:

  1. Determine the right topic directory (create new one if needed)
  2. Write or update the article in docs/topics/<topic>/<article>.md
  3. Update related pages — scan existing articles for references to the same concepts, entities, or data. Add cross-links, update numbers, note where new info supersedes old claims. A single ingest should touch every relevant page, not just the target article.
  4. Update docs/topics/<topic>/_index.md with the new article link
  5. Update docs/index.md stats if a new topic/article was added
  6. Update mkdocs.yml nav section
  7. Append an entry to docs/log.md
  8. Rebuild: scripts/build.sh

Articles should be self-contained, factual, and cross-linked where relevant. Use See also: [Title](relative-path.md) for connections.

2. Compiling from raw/

When the user drops sources in docs/raw/ or gives you a URL:

  1. Read/fetch the source material
  2. Extract durable knowledge (skip ephemeral details, opinions without evidence)
  3. File into appropriate topic articles (new or existing)
  4. Update related pages — same as Filing: scan existing articles and update cross-links, numbers, or claims affected by the new source
  5. Move the raw source to docs/raw/processed/
  6. Append an entry to docs/log.md
  7. Rebuild: scripts/build.sh

3. Linting (daily via heartbeat)

Scan ~/wiki/docs/ for:

  • Contradictions — facts/numbers/claims in one article that conflict with another. Flag the specific pages and the conflicting statements.
  • Stale data — outdated references, old dates paired with "current"/"latest" language, version numbers that have been superseded
  • Missing links — references to topics without articles
  • Dead cross-links — broken See also links
  • Orphan pages — pages with no inbound links from other articles
  • Weak pages — articles that are too thin to be useful on their own (candidates for merging or expanding)

Fix within ~/wiki/docs/: broken links, typos, missing cross-links, orphan pages (add links from related articles). Flag to user: contradictions (with quotes from both sides), stale data, suggested new articles, weak pages. Append a lint entry to docs/log.md.

4. Filing Good Answers

When a conversation produces a solid synthesized answer (comparison, analysis, deep-dive), proactively offer to file it as a wiki page. Good candidates:

  • Comparisons or evaluations (model benchmarks, tool comparisons)
  • How-to knowledge derived from troubleshooting
  • Synthesized research across multiple sources
  • Decisions with rationale

These shouldn't disappear into chat history — they compound in the wiki.

5. Browsing

The wiki is served as a static MkDocs site. After any content change, run scripts/build.sh to regenerate.

Article Style Guide

  • Title: H1 at top, descriptive
  • Date/context: italicized line under title when relevant (e.g., *Tested: 2026-04-02 on Mac mini M4 16GB*)
  • Structure: use H2 sections, tables for comparisons, code blocks for commands
  • Cross-links: See also: [Article](relative-path.md) at bottom
  • Be factual: cite sources, include numbers, avoid vague claims
  • Keep it dense: this is a reference, not a blog post

Adding New Topics

  1. Create docs/topics/<topic>/ directory
  2. Create docs/topics/<topic>/_index.md with overview + article list
  3. Add to docs/index.md topic list
  4. Add to mkdocs.yml nav section

Dependencies & System Changes

This skill requires or will install the following:

DependencyPurposeInstalled by
Python 3MkDocs runtimeMust be pre-installed
mkdocs + mkdocs-materialStatic site generatorbootstrap.sh via pipx or pip3
gitVersion controlMust be pre-installed

What bootstrap.sh does

  • Creates ~/wiki/ directory structure
  • Installs mkdocs + mkdocs-material via pipx/pip3 if not already installed
  • Initializes a git repository
  • Creates a persistent static server as a LaunchAgent (macOS) or systemd user service (Linux) on port 8300
  • The static server serves ~/wiki/site/ on 127.0.0.1 only (not publicly accessible)

What build.sh does

  • Rebuilds the MkDocs static site
  • Commits all changes to git (local only by default)
  • Use --push to also push to a configured git remote (requires SSH keys or stored HTTP credentials)

Optional: heartbeat integration

You can integrate wiki linting into your agent's heartbeat cycle. This is opt-in and requires adding config to your HEARTBEAT.md. The heartbeat integration reads heartbeat-state.json (for timing) and optionally memory/ daily notes (to detect wiki gaps). See references/heartbeat-integration.md for setup.

Log (docs/log.md)

Append-only chronological record of wiki activity. Every ingest, lint, and filed answer gets an entry. Format:

## [YYYY-MM-DD] ingest | Article Title
- Source: <URL or filename>
- Pages touched: page1.md, page2.md, ...
- Summary: one-line description of what was added/changed

## [YYYY-MM-DD] lint
- Fixed: <list of silent fixes>
- Flagged: <list of issues reported to user>

## [YYYY-MM-DD] filed | Article Title
- Origin: conversation / Q&A synthesis
- Summary: one-line description

The log is parseable with grep: grep "^## \[" docs/log.md | tail -10 gives the last 10 entries.

Build & Deploy

Always run scripts/build.sh after content changes. This:

  1. Rebuilds MkDocs static site
  2. Commits all changes to git
  3. Pushes to remote

平台分布

OpenClaw

77.92%
按下载量换算7,362

安全审计

VirusTotal

未展示

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