ReadGZH

让 AI 读懂微信公众号。自研 7 阶段提取管线，99.89% 穿透反爬，Token 消耗降低 50–87%。支持 ChatGPT、Claude、Perplexity、Gemini 等平台无缝引用。| Let AI read and understand WeChat Official Accounts. Sel...

Before installing, consider that the skill will send URLs and article content to https://readgzh.site (their API/MCP server) and that the package includes an embedded API key (credentials.json). That embedded key is unexpected and risky: it could be abused, may be a shared/demo key, and could expose your activity or cause billing/abuse on the owner's account. Ask the publisher to remove the credentials file (or provide clear justification), rotate the exposed key, and instead require users to supply their own API key via a documented env var. If you don't trust readgzh.site, do not configure MCP or use the bundled key; use your own API key and review privacy/retention terms (shared cache means read articles may be stored and available to other users). If you proceed, monitor usage of any API key you supply and avoid passing sensitive or private article URLs you don't want stored on a third-party cache.

Type: OpenClaw Skill Name: readgzh Version: 1.3.9 The skill bundle provides a functional tool for parsing WeChat articles into Markdown via the ReadGZH API (api.readgzh.site). It is classified as suspicious primarily due to the inclusion of a hardcoded 'live' API key in credentials.json, which represents a security vulnerability (credential exposure). While the instructions in SKILL.md and the OpenAPI specification are consistent with the stated purpose, the presence of pre-configured credentials in a distributed bundle is a high-risk practice.