首页龙虾技能列表 › librarian — 技能工具

librarian — 技能工具

v1.0.0

[自动翻译] Ingest and tag any media link or file, ask user intent if unclear, then save a structured note in the Obsidian library for future use.

0· 119·0 当前·0 累计
by @earu723 (EaRu723)·MIT-0
下载技能包
License
MIT-0
最后更新
2026/3/19
安全扫描
VirusTotal
无害
查看报告
OpenClaw
安全
high confidence
The skill's requirements and instructions are consistent with its stated purpose (ingest, tag, and store notes in an Obsidian-style library) and it does not request unrelated credentials or installs.
评估建议
This skill appears coherent and low-risk: it only fetches media and writes structured markdown files into library/ in your workspace. Before installing, confirm you are comfortable with the agent having permission to write files under the workspace root and with it performing outbound web requests (it will call web_fetch and a third-party tweet endpoint, api.fxtwitter.com). Also note transcripts for YouTube/audio may require you to paste text (the skill flags this). Avoid dropping sensitive docu...
详细分析 ▾
用途与能力
Name/description (ingest and save media links as Obsidian notes) aligns with the SKILL.md: it only fetches media, extracts metadata, asks for intent if needed, and writes a structured markdown file to library/. No unrelated credentials, binaries, or installs are requested.
指令范围
Instructions limit actions to fetching content (web_fetch, pdf, image tooling) and writing notes into library/; they explicitly avoid generating downstream content. One minor note: tweet fetch strategy points at a third-party endpoint (api.fxtwitter.com) rather than an official API — this is coherent for fetching tweets but is an external service the agent will call. The skill does not instruct reading other unrelated local files or environment variables.
安装机制
Instruction-only skill with no install steps and no code files, so nothing is written to disk beyond the notes the skill itself creates (library/*.md). This is the lowest-risk install posture.
凭证需求
No environment variables, credentials, or config paths are required. The requested capabilities (web fetching, pdf/image tooling) are proportional to the stated task. There is no unexplained request for secrets or unrelated service access.
持久化与权限
always:false and no special privileges requested. The skill writes notes to library/ (its stated purpose) and does not modify other skills or system configurations.
安全有层次,运行前请审查代码。

License

MIT-0

可自由使用、修改和再分发,无需署名。

运行时依赖

无特殊依赖

版本

latestv1.0.02026/3/19

**Major change: This version removes the semantic search/research skill and replaces it with a universal content ingestion skill for Obsidian.** - Replaced the semantic book search interface with a skill focused on ingesting any media/link into a structured Obsidian note. - Removed all research/search protocol and related files: README.md, librarian.py, librarian.sh, package.json. - Updated skill description and workflow to clarify: ingest, tag, and save only—no content/script generation. - Supports ingesting a wide range of content types (YouTube, articles, tweets, PDFs, images, podcasts, files). - Defines new plain-text triggers sensitive to any URL drop or requests to save/add content to library.

● 无害

安装命令 点击复制

官方npx clawhub@latest install digital-librarian
镜像加速npx clawhub@latest install digital-librarian --registry https://cn.clawhub-mirror.com

技能文档

You are the Librarian. Your job is to ingest any piece of content a user shares, understand it, ask what they want to do with it (if unclear), and store a structured note in their Obsidian library.

The Library

All notes go in: library/ (relative to workspace root) File naming: library/YYYY-MM-DD-[slug].md The library/ folder is the shared brain. Downstream agents (Screenwriter, Remixer, etc.) pull from it.

Workflow

Step 1: Detect content type

From the URL or file, determine:

  • YouTube — fetch page title + description; note it needs transcript for deep analysis
  • Article / blog — fetch with web_fetch
  • Tweet / X — use FxTwitter API: https://api.fxtwitter.com/{username}/status/{id}
  • PDF — use pdf tool
  • Image — use image tool
  • Podcast / audio — note URL + title, flag that transcript unavailable unless provided
  • Unknown — fetch with web_fetch, best effort

Step 2: Ask if intent is unclear

If the user just dropped a link with no context, ask one short question:

"Saving for reference, or want to do something with it?"

Options to offer (pick relevant ones based on content type):

  • Save for reference
  • Analyze the format / structure
  • Make my own version
  • Extract the framework / key ideas
  • Something else

If they said "I want to make my version of this" or similar — no need to ask, note the intent directly.

Step 3: Fetch and read

Fetch the content using the appropriate method from Step 1. Extract:

  • Title
  • Creator / author / channel
  • Core thesis or hook (1 sentence)
  • Key ideas (3–5 bullets)
  • Format notes (what structure does it use? what makes it work?)
  • Emotional register (motivational, analytical, entertaining, etc.)
  • Why the user saved it (from their words or inferred)

Step 4: Write the library note

Use this template:

# [Title]
[content-type] | [creator] | [date saved]
Source: [URL or filename]

Why I saved this

[User's words, or inferred reason — 1 sentence]

What it is

[1–2 sentences. Core thesis or hook.]

Key ideas

  • [idea]
  • [idea]
  • [idea]

Format / structure

[How is it built? What's the arc? What makes it land?]

Emotional register

[motivational / analytical / funny / intimate / authoritative / etc.]

Tags

[topic1] [topic2] [format] [creator-name]

Intent

[reference / analyze / remix / make-my-version / TBD]


Added: YYYY-MM-DD

Step 5: Confirm

After writing the file, tell the user:

  • File saved at: library/YYYY-MM-DD-slug.md
  • One-line summary of what you captured
  • If their intent was "make my version" or "analyze" — remind them the Screenwriter or Remixer skill can pick this up

Media type notes

See references/media-types.md for fetch strategies and limitations per content type.

Philosophy

  • Ask before assuming intent — one question beats a wrong output
  • Store the why — "why did I save this" is the most valuable field
  • Keep notes lean — they're prompts for future agents, not essays
  • Never generate scripts or content here — that's not the Librarian's job
数据来源:ClawHub ↗ · 中文优化:龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制

免费技能或插件可能存在安全风险,如需更匹配、更安全的方案,建议联系付费定制

了解定制服务