Obsidian Clipper

Save web content (articles, videos, notes) to Obsidian vault with automatic classification, intelligent naming, and content extraction. Supports: 小红书 (Xiaoho...

Security Scan

The skill generally matches its stated purpose (save web content to an Obsidian vault), but there are several inconsistencies and a shipped config that could cause unexpected writes or privacy leakage — review config and file-write behavior before installing.

Purpose & Capability

The skill's declared purpose (clip web content to an Obsidian vault) aligns with the SKILL.md instructions (fetch content, format Markdown, write to vault). However README and SKILL.md reference external tools (obsidian-cli, browser snapshots, web_fetch/web_search/write tools) while the registry metadata declares no required binaries — a mild inconsistency. Shipping a pre-filled config.json pointing to the author's personal vault path is unexpected for a general-purpose skill.

Instruction Scope

Runtime instructions tell the agent to read/create a config.json from the skill directory and to write files to the filesystem (vault_path). Because a config.json is included in the package, the skill would use that existing config without prompting the user, potentially attempting writes to the hardcoded path. Instructions also require web_fetch/web_search/browser snapshot capabilities; they don't direct data to external endpoints outside normal web fetching, but the silent/automatic saving behavior and reading/writing of local paths merit caution.

Install Mechanism

This is an instruction-only skill with no install spec or code to download/execute. That is the lowest install risk. README suggests optional install methods (clawhub/git clone) but nothing in the package executes arbitrary downloads at install time.

Credentials

The skill requests no environment variables or credentials (good), but the included config.json hardcodes an absolute vault_path and a collector_name (author's iCloud/Obsidian path). Bundling a user-specific absolute path is disproportionate/unexpected and could cause the skill to operate on an unintended filesystem location or leak the author's usage pattern. The SKILL.md also references tools (obsidian-cli, browser) not declared in metadata.

Persistence & Privilege

The skill does not request always:true or other elevated privileges. It will create/modify config.json in its own directory and write Markdown files into the user-specified vault_path; this is expected for its purpose. Be aware the skill is allowed to run autonomously by default (disable-model-invocation: false), which combined with silent auto-save behavior could result in files being written without follow-up confirmation unless the platform enforces user prompts on skill invocation.

What to consider before installing

Before installing or using this skill: - Inspect and edit the included config.json immediately. It ships with an absolute vault_path and collector_name belonging to the author; change the vault_path to your intended location or delete the file so the skill must prompt you on first run. - Understand that the skill writes Markdown files to whatever vault_path is present in config.json — test with a disposable folder first to confirm behavior. - Ensure the platform 'write' and 'web_fetch'/'web_search' tools it expects are the ones you trust; the SKILL.md also mentions browser snapshot and obsidian-cli (not declared in metadata). Install obsidian-cli only if you need CLI integration. - Because classification is silent by default, consider instructing the skill to prompt you for classification or review saved items until you’re comfortable with automatic behavior. - If you are concerned about autonomous invocation, consider disabling autonomous skill invocation or only enabling this skill when explicitly invoked. These steps reduce the risk of unexpected writes or privacy issues stemming from the packaged configuration and silent-save behavior.

将网页、文章、视频、小红书笔记等内容收藏到 Obsidian vault。

配置

Skill 首次运行时，必须先确认以下配置。读取 config.json（与本 SKILL.md 同目录），不存在则询问用户并创建。

config.json 格式

{
  "vault_path": "/path/to/your/vault/收藏文档/",
  "collector_name": "你的名字",
  "categories": {
    "AI工具": ["AI", "机器学习", "LLM", "GPT", "Claude", "OpenClaw", "Agent", "量化"],
    "兴趣爱好": ["路亚", "钓鱼", "户外", "游戏", "音乐", "运动", "3D打印"],
    "技术教程": ["编程", "代码", "教程", "开发", "配置", "部署", "API", "DevOps"],
    "生活日常": ["美食", "旅行", "健康", "生活技巧", "读书"]
  }
}

配置项说明

字段	必填	说明
vault_path	✅	Obsidian Vault 中收藏文档的根目录路径（绝对路径，末尾带 /）
collector_name	❌	签名用的名字，默认不签名
categories	❌	自动分类规则，键=目录名，值=关键词数组。默认包含常用分类（见示例）

首次运行流程

1. 检查 config.json 是否存在
2. 不存在 → 询问用户 Vault 路径，分类使用默认值
3. 创建 config.json 保存配置
4. 后续运行直接读取，用户可随时要求修改

用户修改配置

用户随时可以说：

• "把 Vault 路径改成 xxx"
• "加一个「读书笔记」分类"
• "我的名字改成 xxx" → 更新 config.json

文件命名规范

格式: 「来源类型」主题-YYYY.MM.DD.md

来源类型:

• 「小红书」 — 小红书笔记
• 「Youtube视频」 — YouTube 视频
• 「B站视频」 — B站视频
• 「知乎文章」 — 知乎文章
• 「公众号」 — 微信公众号文章
• 「网页」 — 普通网页
• 「技术文档」 — 官方文档、技术博客
• 「短视频」 — 抖音/快手等短视频

标题提取策略

按优先级获取标题：

1. 页面 <title> 标签 — 最可靠
2. og:title meta 标签 — 社交平台常用
3. 搜索结果中的标题 — web_search 返回的 title
4. 用户消息中的描述 — 用户发送时的上下文

原则：标题简洁、准确。过长时（>50字）截取核心部分。

工作流程

1. 接收输入

用户可能提供：

• 直接链接（小红书、YouTube、知乎、公众号等）
• 内容描述 + 要求收藏
• 多个链接批量处理
• 截图 + 要求收藏

2. 获取内容

根据来源类型选择方法（三层降级）：

第一层：web_fetch 直接抓取
  ↓ 失败
第二层：web_search 搜索相关内容，整合补充
  ↓ 搜索无结果
第三层：浏览器打开页面抓取（browser snapshot）
  ↓ 仍失败
提示用户提供内容（复制文字/截图）

各平台策略

平台	识别特征	策略
普通网页/知乎	通用 URL	web_fetch → web_search
小红书	xhslink.com	⚠️ 登录墙，web_search 降级
微信公众号	mp.weixin.qq.com	⚠️ 必须用浏览器抓取
抖音/快手	v.douyin.com / douyin.com	❌ 无法解析，提示用户截图/复制文字
YouTube	youtube.com / youtu.be	web_search 搜索标题 + 要点
B站	bilibili.com / b23.tv	web_fetch → web_search

3. 自动分类

读取 config.json 中的 categories 规则，默认自动分类，不打扰用户：

1. 用户明确指定 → 直接使用，无视关键词匹配
2. 关键词匹配 → 遍历 categories，内容包含关键词最多的分类胜出
3. 无匹配 → 归入第一个分类，不询问
4. 用户说"放到 XX" → 直接创建 vault_path/XX/ 目录

原则：静默自动分类。只在用户主动要求时才讨论分类。

4. 生成文档

Markdown 模板:

# {标题}

**来源**: {来源链接或描述}
**保存日期**: {YYYY.MM.DD}
**类别**: {分类名}

---

## 摘要

{简短摘要，1-3句话}

---

## 正文内容

{整理后的内容}

---

## 关键要点

- {要点1}
- {要点2}
- {要点3}

---

## 相关链接

- {原始链接}
- {其他参考}

---
{如果 config.json 中设置了 collector_name，追加：*收藏人: {collector_name}*}

5. 保存文件

1. 从 config.json 读取 vault_path + 确定分类
2. 确定文件名
3. 使用 write 工具保存（自动创建目录）
4. 回复用户：✅ 已保存 + 路径 + 核心要点

批量收藏

用户一次发送多个链接时，逐个处理，统一回复：

✅ 已保存 3 篇到 Obsidian

1. 「小红书」AI转3D工具 → AI工具/
2. 「小红书」路亚教程 → 兴趣爱好/
3. 「小红书」OpenClaw优化 → AI工具/

注意事项

1. 首次运行必须配置 — 读取或创建 config.json
2. 小红书/抖音 — 有登录墙，降级到搜索或提示用户提供内容
3. 公众号文章 — 必须用浏览器抓取
4. 默认自动分类 — 静默完成，不打扰用户。无匹配时归入第一个分类
5. 内容过长 — 提取核心要点
6. 日期格式 — 统一 YYYY.MM.DD
7. 浏览器用完关闭 — browser open 后记得 close
8. 不要硬编码个人路径/名字 — 一切从 config.json 读取