Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

AI Info Digest

v1.0.0

微信公众号文章摘要整理工具。获取用户关注的微信公众号在指定时间范围内发布的文章,逐篇阅读全文,然后按主题聚合整理成一篇简明的摘要文章,标注来源便于评估各公众号的价值。 当用户提到以下场景时触发:整理公众号文章、公众号周报/日报、微信文章摘要、帮我看看公众号最近发了什么、总结一下这几天的公众号内容、公众号阅读整理、...

0· 79·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for jinntrance/ai-info-digest.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "AI Info Digest" (jinntrance/ai-info-digest) from ClawHub.
Skill page: https://clawhub.ai/jinntrance/ai-info-digest
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install ai-info-digest

ClawHub CLI

Package manager switcher

npx clawhub@latest install ai-info-digest
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The name/description (WeChat/public-account digest) aligns with the runtime instructions: the SKILL.md explains how to locate articles on Zhihu, official sites, or via web search and then summarize them. The included wechat-accounts.json matches the stated purpose and supplies example accounts. Requiring a browser-reading tool (Claude in Chrome MCP) is reasonable for extracting web article text.
!
Instruction Scope
The SKILL.md tells the agent to read a configuration file from the user's home (~/.claude/skills/wechat-digest/wechat-accounts.json) and to save digests to ~/Documents/ai-info-digest/{date}_digest.md. The manifest metadata claimed 'Required config paths: none', so the runtime instructions reference local file I/O that is not declared in the metadata. The skill also instructs the agent to navigate arbitrary web pages and extract full page text (including Zhihu and other sites). These behaviors are coherent with the feature but broaden the scope (reading/writing files and browsing with potential access to logged-in sessions).
Install Mechanism
This is an instruction-only skill with no install spec and no code files to execute. That minimizes install-time risk (nothing is downloaded or installed by the skill).
!
Credentials
No environment variables or credentials are requested, which is appropriate. However, the skill requires the 'Claude in Chrome' MCP browsing tool and will navigate and extract page text. That tool may leverage browser state (cookies, authenticated sessions); consequently the agent could read pages that require login using the user's browser session. The skill also reads/writes files in the user's home directory (config + saved digest). Both are proportionate to the task but are privacy-sensitive and were not reflected in the declared config-path metadata.
Persistence & Privilege
The skill is not always-enabled and does not request elevated platform privileges. It will write output files to the user's Documents folder and expects a per-skill config file under ~/.claude/skills/..., which is normal for an instruction-only skill but constitutes persistent files on disk. Autonomous invocation is allowed (default) — combine that with browsing access if you are concerned about automatic runs.
What to consider before installing
This skill mostly does what its description promises (collecting and summarizing WeChat/public-account content), but check a few things before installing: - The SKILL.md expects a local config file at ~/.claude/skills/wechat-digest/wechat-accounts.json and writes digests to ~/Documents/ai-info-digest/. The registry metadata did not declare these config paths — if you care about filesystem changes, inspect or create that config file yourself and verify where outputs will be saved. - The skill requires a browser-reading tool (Claude in Chrome MCP) which can access web pages using your browser session. If you are logged into sites (Zhihu, WeChat, etc.), the tool may read pages that require authentication — consider running it only when you are comfortable with that access or using a separate browser profile. - There are no secrets requested, and no installers or remote-download steps, which reduces supply-chain risk. Still, review the wechat-accounts.json to remove any private entries and confirm the output path before running. If you want to be extra cautious: create the expected config path yourself with a minimal accounts list, run the skill in a controlled environment/profile, and verify the files it writes.

Like a lobster shell, security has layers — review code before you run it.

aivk9718e5zaga5dwn4kvrxtj20qh84d3ahdigestvk9718e5zaga5dwn4kvrxtj20qh84d3ahlatestvk9718e5zaga5dwn4kvrxtj20qh84d3ahnewsvk9718e5zaga5dwn4kvrxtj20qh84d3ahwechatvk9718e5zaga5dwn4kvrxtj20qh84d3ah
79downloads
0stars
1versions
Updated 3w ago
v1.0.0
MIT-0

微信公众号文章摘要整理

概述

收集指定时间范围内用户关注的微信公众号文章,逐篇阅读后按主题聚合整理成简明摘要。每条摘要标注来源公众号,便于用户评估各号的价值(是否值得继续关注)。

前置条件

  • Claude in Chrome MCP 工具(用于浏览器阅读知乎/官网文章全文)
  • WebSearch 工具(用于搜索兜底)
  • 公众号列表配置文件:~/.claude/skills/wechat-digest/wechat-accounts.json

配置文件结构:

{
  "accounts": [
    {
      "name": "量子位",
      "zhihu": "https://zhuanlan.zhihu.com/qbitai",
      "website": "qbitai.com",
      "priority": "high"
    }
  ]
}
  • zhihu:知乎专栏或个人主页链接(首选获取渠道,绝大多数公众号在此同步发布)
  • website:公众号独立网站(次选)
  • priority:high 的公众号文章摘要更详细

各公众号知乎链接速查

公众号知乎链接类型
新智元zhihu.com/org/xin-zhi-yuan-88-3/posts机构主页(文章列表)
创业邦zhuanlan.zhihu.com/chuangyebang机构专栏
运筹OR帷幄zhuanlan.zhihu.com/operations-research机构专栏
刘聪NLPzhihu.com/people/LiuCongNLP个人主页
量子位zhuanlan.zhihu.com/qbitai机构专栏
机器之心zhuanlan.zhihu.com/jiqizhixin机构专栏
集智俱乐部zhuanlan.zhihu.com/swarma机构专栏
数字生命卡兹克zhihu.com/people/Khazix个人主页
PaperWeeklyzhuanlan.zhihu.com/paperweekly机构专栏
AI科技评论zhuanlan.zhihu.com/leiphone机构专栏
AI前线zhuanlan.zhihu.com/infoq机构专栏
AINLP(无知乎,用WebSearch兜底)

工作流程

第一步:确认输入

  1. 时间范围:用户提供日期范围(如 "最近一天"、"4月1日到4月3日"),根据当前日期计算具体起止日期。
  2. 公众号列表:从 ~/.claude/skills/wechat-digest/wechat-accounts.json 读取。

第二步:获取文章列表

对每个公众号,按以下优先级获取时间范围内的文章:


方案 A(首选):知乎专栏 / 个人主页

知乎是绝大多数公众号的内容镜像站,无访问限制,文章时间戳精确。

机构专栏zhuanlan.zhihu.com/xxx):

  1. 导航到专栏文章列表:https://zhuanlan.zhihu.com/xxx?order_by=hot(或直接访问专栏主页)
  2. get_page_text 提取文章列表,找到时间范围内的文章标题 + 链接
  3. 逐篇导航并 get_page_text 读取全文

机构主页zhihu.com/org/xxx/posts):

  1. 直接导航到 /posts 页面(如新智元:https://www.zhihu.com/org/xin-zhi-yuan-88-3/posts
  2. 文章按时间倒序排列,get_page_text 读取列表
  3. 逐篇导航并读取全文

个人主页zhihu.com/people/xxx):

  1. 导航到 https://www.zhihu.com/people/xxx/posts(文章列表)
  2. 提取时间范围内的文章
  3. 逐篇读取全文

知乎日期识别技巧

  • 知乎文章时间显示为"X 天前"、"X 小时前"或具体日期
  • 根据当前日期反推,确认是否在目标时间范围内
  • 如果列表按时间倒序排列,看到超出范围的文章即可停止

方案 B(次选):公众号官方网站

适用于有独立网站且内容同步的公众号(量子位→qbitai.com,机器之心→jiqizhixin.com 等):

  1. 导航到官网首页
  2. 提取最新文章列表(通常按时间倒序)
  3. 筛选时间范围内的文章并逐篇读取

方案 C(兜底):WebSearch

适用于知乎和官网都无法获取的情况(如 AINLP):

  • 搜索 "{公众号名称}" site:mp.weixin.qq.com {YYYY}年{M}月
  • 或搜索 "{公众号名称}" 最新 {YYYY}年{M}月{D}日
  • 注意:搜索结果日期可能不精准,需人工核实

执行策略

  • 优先用知乎批量处理所有有链接的公众号
  • 可以在同一个标签页依次访问,无需开多个标签
  • 有独立网站的公众号失败后才用官网
  • 某公众号所有方案都失败时,记录到"未能获取"列表,不阻塞其他公众号的处理

第三步:阅读文章全文

对每篇文章:

  1. mcp__Claude_in_Chrome__navigate 打开链接
  2. mcp__Claude_in_Chrome__get_page_text 提取全文
  3. 记录:公众号名称、标题、发布日期、核心要点

注意mp.weixin.qq.com 链接通常被浏览器安全策略拦截,优先使用知乎或官网的同步文章链接。

第四步:整理摘要

# 公众号摘要:{日期范围}

> {N} 个公众号 · {M} 篇文章

## 要点速览
- 要点1 `[来源号]`
- 要点2 `[来源号]`

## 按主题整理

### 主题一

- **文章标题** `[公众号名]` — 1-2句核心内容,关键数据/结论。
- **文章标题** `[公众号名]` — ...

### 主题二
...

## 各公众号本期表现

| 公众号 | 文章数 | 信息密度 | 独家/原创 | 建议 |
|--------|--------|----------|-----------|------|
| xx     | 3      | ★★★★   | 有        | 保留 |
| yy     | 5      | ★★☆☆   | 无,多转载 | 可考虑取关 |

## 未能获取的公众号
- AINLP:知乎暂无专栏,WebSearch未找到近期文章

## 原文链接
1. [文章标题](链接) — 公众号名

整理原则

  • 按主题聚合,不按公众号分组;同话题多来源观点并列
  • 每篇摘要控制在 1-2 句,提炼信息增量,不复述
  • [公众号名] 标注来源
  • 评估表帮用户判断哪些号值得保留:信息密度标准是独家内容、数据质量、观点深度,而非发文量

第五步:输出

  1. 保存到 ~/Documents/ai-info-digest/{YYYY-MM-DD}_digest.md
  2. 对话中展示完整摘要
  3. 告知文件保存路径

异常处理

情况处理方式
知乎专栏显示"X天前"日期模糊根据当前日期推算,宁可多取不漏
知乎需要登录才能看文章切换方案B(官网)或方案C(WebSearch)
官网文章列表加载依赖JSread_page 替代 get_page_text,查找文章链接
某公众号近期无更新在评估表中标注"本期无发布"
验证码 / 反爬触发立即停下通知用户,等待人工干预

Comments

Loading comments...