Token导航 LogoToken导航

Agent Skill · 研究检索

vault-scribe(保险库抄写员)

vault-scribe 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 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。

vault-scribe 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为操作浏览器;许可证信息还不明确,第三方扫描有未通过项,按安装量在研究检索类的 73,675 个 Skill 里排第 60,816。

总安装 364周安装 15GitHub Stars 公开资料未说明下载 119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
unknown

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:vault-scribe(保险库抄写员)
来源仓库:https://github.com/psenger/ai-agent-skills
仓库路径:skills/vault-scribe
安装命令:
npx skills add https://github.com/psenger/ai-agent-skills --skill vault-scribe
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 npx skills 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。

skills.shnpx skills
npx skills add https://github.com/psenger/ai-agent-skills --skill vault-scribe

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。

vault-scribe 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、任务场景或来源线索快速定位候选结果。它可辅助从文档中提取内容、匹配字段或过滤无关信息,提升信息获取效率。

通过 npx skills add 命令从指定仓库安装,具体用法需结合 README 进一步确认。安装前建议检查权限范围和维护状态,避免触发不必要的联网或文件操作。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/vault-scribe

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 Codex、Claude、Cursor、Gemini CLI;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 npx skills add https://github.com/psenger/ai-agent-skills --skill vault-scribe;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

    它需要操作浏览器,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

vault-scribe 怎么安装?

在终端执行 npx skills add https://github.com/psenger/ai-agent-skills --skill vault-scribe。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

vault-scribe 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。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。

vault-scribe 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

vault-scribe 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

vault-scribe 用的人多吗?

公开数据是累计安装 364 次,近一周安装 15 次,下载 119 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。

vault-scribe 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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 Codex,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。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 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
vault-scribe

保险库抄写员 · 本页

364—操作浏览器未明确未通过

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

Vault Scribe — Obsidian + GitHub Markdown Skill

Creates and edits Markdown that is GitHub-Flavored Markdown (GFM) first, with optional Obsidian-specific extensions when needed. Standard Markdown (headings, bold, italic, lists, quotes, code blocks, tables) is assumed knowledge.

Default rule: Use standard GFM syntax unless the user explicitly requests Obsidian-specific features or is working in an Obsidian-only context. Obsidian wikilinks ([[Note]]) break on GitHub — always prefer standard Markdown links for cross-platform compatibility.

When invoked with an argument (e.g. /vault-scribe meeting), use $ARGUMENTS to determine the note type. If no argument is provided, infer the type from the source material or ask the user.


Workflow: Creating a Note

1. Analyse the Source Material

  • Read the full transcript, notes, or brief carefully
  • Identify the core topic, intended audience, and key concepts
  • Note any implicit structure (e.g., problem → solution → examples)
  • Extract any explicit examples, warnings, tips, or quotes worth preserving

2. Generate YAML Front Matter

Reference: See references/FRONT-MATTER.md for the complete schema definitions, all enum values, and type-specific required fields.

First, determine the note type based on what the user is creating:

If the user wants…Set type to
A guide, reference doc, or knowledge articlearticle
A step-by-step instructional guide or procedurehow-to
Architecture docs, RFCs, design docs, system specstechnical
An investigation with multiple sourcesdeep-research
A versioned plan or strategy documentstrategy
Meeting notes, 1:1s, standups, retrospectivesmeeting
Brainstorming, ideation, or exploratory thinking (solo or group)brainstorming
[!TIP] Brainstorming is its own note type. It works for group sessions (add attendees and meeting_date) and solo AI ideation (omit them). Add sources when the brainstorm draws on external references.

Then apply the corresponding frontmatter schema from references/FRONT-MATTER.md. Every note type uses the core fields plus any type-specific required fields.

Tag guidelines:

  • Use lowercase, hyphenated slugs only (e.g. distributed-systems, not Distributed Systems)
  • Include 4–8 tags: topic-specific first, then broader domain tags
  • Always include at least one technology tag and one domain/concept tag

3. Structure the Document Body

Use this hierarchy:

# Title (H1) — matches front matter title
## Section (H2) — major topic areas
### Subsection (H3) — specific concepts within a section

Target sections (adapt names to the topic):

  1. Overview / What is X? — define the subject clearly
  2. How It Works — mechanics, process, or architecture
  3. Practical Examples — concrete code blocks or walkthroughs
  4. Common Mistakes / Warnings — callout blocks
  5. Quick-Start / Checklist — actionable summary
  6. Further Reading & References — links table
[!NOTE] Minimum 4 sections (Overview, How It Works, Examples, References) for all notes. The full 6 sections are the target for comprehensive articles and guides. Simpler note types (meetings) naturally use fewer sections.

4. Apply Callout Blocks

Use GFM Alerts by default (renders on both GitHub and Obsidian). See references/CALLOUTS.md for the full list of callout types including Obsidian-only variants.

Callout TypeUse For
[!NOTE]Neutral supplementary info
[!TIP]Actionable best practice
[!IMPORTANT]Key concept the reader must not miss
[!WARNING]Common mistake or gotcha
[!CAUTION]Risk of data loss, security issue, or breaking change
[!TIP] Every Warning, Tip, and TL;DR should be a callout block — not plain prose. This makes the document scannable.

5. Code Blocks

Always use fenced code blocks with a language identifier:

key: value

mkdir -p ~/.claude/skills/my-skill

For directory trees, use plain text or no language tag.

6. Table of Contents

Never use [TOC], [[_TOC_]], or any other TOC directive — neither GitHub nor Obsidian supports them natively. They render as broken plain text.

  • GitHub auto-generates a TOC in the sidebar — no directive needed.
  • Obsidian generates one via its built-in plugin — no directive needed.
  • Only add a manual TOC (using standard anchor links) when the document has 6+ sections and will be read outside a browser. See references/MARKDOWN-SYNTAX.md for anchor rules and an example.

7. Tables & Links

Use Markdown tables for comparisons, option lists, and reference links. Always include a reference links table at the end:

| Resource | Link |
|---|---|
| Official Docs | [docs.example.com](https://docs.example.com) |
| Source Video  | [youtube.com/watch?v=...](https://youtube.com/watch?v=...) |

8. Inline Formatting Rules

ElementUsage
**bold**Key terms on first use, critical values
*italic*Titles of external resources, emphasis
` code `All file paths, commands, config keys, code symbols
[[wikilink]]Internal Obsidian links (only if Obsidian-only context)

9. Transcript Appendix

When the source material includes a transcript (video, podcast, meeting recording, article), always append the raw transcript at the very end of the document, after all other content, using this exact format:

---

## Transcript

Transcript from: [<name of video or article or meeting>](<link to source>)
Date of material: <YYYY-MM-DD>

\```
<raw transcript>
\```
  • The transcript section is separated from the rest of the document by a horizontal rule (---)
  • The source name must be a clickable Markdown link when a URL is available
  • If no URL exists, use plain text for the name
  • Date should be in YYYY-MM-DD format when known; leave blank if unknown
  • The raw transcript goes inside a fenced code block with no language tag, preserving the original text exactly as provided

Output Requirements

  • File extension: .md
  • Encoding: UTF-8
  • Front matter: Always present, always first
  • Minimum sections: 4 (Overview, How It Works, Examples, References). Target 6 for comprehensive articles.
  • All external URLs must be real and formatted as [label](url)
  • No bare URLs — always wrapped in Markdown link syntax
  • End the document with an [!abstract] callout containing the TL;DR (before the Transcript appendix, if present)

Quality Checklist

Before finalising the output, verify:

  • Front matter is complete and valid YAML
  • type field is set and matches the correct note type schema
  • All type-specific required fields are present (per references/FRONT-MATTER.md)
  • Title in front matter matches H1 heading
  • category is a valid enum value
  • Tags are lowercase and hyphenated
  • All warnings/tips are callout blocks, not plain paragraphs
  • At least one code block example is present (for articles/guides)
  • A reference links table exists at the end
  • Document closes with an [!abstract] callout (before Transcript appendix if present)
  • No bare URLs (all links use [label](url) format)
  • date_created and date_updated are set (use today's date if unknown)
  • If source material includes a transcript, it is appended in the Transcript Appendix format

Reference Files

  • references/FRONT-MATTER.md — Complete frontmatter schemas for all note types, enum values for type, category, and status, and type-specific required/optional fields. Always consult this file when generating front matter.
  • references/CALLOUTS.md — GFM Alerts (default) and Obsidian callout types, foldable/nested syntax, aliases, and custom CSS callouts.
  • references/EMBEDS.md — Standard GFM image syntax (default) and Obsidian-specific embed syntax for notes, images, audio, and PDFs.
  • references/MARKDOWN-SYNTAX.md — Detailed GFM + Obsidian syntax reference for links, tags, comments, highlighting, math, diagrams, footnotes, and the compatibility matrix.
  • examples/ — Example output files showing correctly formatted notes for different note types.

References

平台分布

Codex

34%
按下载量换算40

Claude

33.5%
按下载量换算40

Cursor

20.12%
按下载量换算24

Gemini CLI

9.04%
按下载量换算11

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

未通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。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