wx-mp-push

自动化微信公众号文章发布，通过 API 创建和管理文章草稿，支持 Markdown 和 HTML 两种输入格式，自动 MD→HTML 转换，支持封面图和正文图片自动上传。用于：发布文章到公众号、创建内容草稿、自动化内容工作流、或设置定时发布系统。

This skill appears to implement the advertised WeChat article-publishing features, but proceed with caution: 1) It expects you to put AppID/AppSecret into config.json — keep that file private and do not check it into source control. 2) The publish script will scan article HTML for <img src="..."> and will open local files (relative or absolute) and upload them to WeChat; if the HTML contains paths to sensitive files (e.g., /etc/… or other local secrets) those files could be read and sent to an external service. Don't run this on machines containing sensitive data unless you inspect the code and control the content it processes. 3) Documentation has contradictions about handling external URLs — verify whether external remote images are fetched and re-uploaded. 4) Review the script (publish_article.py) yourself, run it in an isolated environment (or with a test account) first, and restrict where config.json and the token cache are stored. If you want a safer setup, provide only sanitized content (no absolute paths), or run the tool in a container with limited filesystem visibility and network access. If you want, I can point out specific lines in publish_article.py that perform file reads and uploads so you can inspect them more closely.

Type: OpenClaw Skill Name: wx-mp-push Version: 1.0.0 The skill is a functional WeChat article publisher but contains a Server-Side Request Forgery (SSRF) vulnerability in 'scripts/publish_article.py'. The '_download_url_to_temp' function fetches content from arbitrary URLs provided in the article content without validation, which could be exploited to access internal network resources. Additionally, the script handles sensitive WeChat credentials (AppID/AppSecret) and stores access tokens locally in plaintext, which are risky but necessary capabilities for its stated purpose. No evidence of intentional malice or data exfiltration was found.