小红书 MCP 服务

小红书（XHS/RED）自动化助手。完整的小红书操作能力，包含 MCP 服务端。 当用户提到小红书、红书、XHS、RED、发笔记、搜笔记、小红书运营等任何与小红书相关的操作时使用此技能。

This skill appears to implement the stated Xiaohongshu automation features, but it ships with data/cookies.json that contains session tokens (access-token-*, id_token, web_session, etc.). Those are effectively credentials and let anyone running the skill act as that account. Before installing or running: 1) Do not run the service with the bundled cookies.json; inspect data/cookies.json and remove or replace it with your own cookies only after you understand the contents. 2) Prefer to run 'npm run login' yourself (the provided login.js opens a browser for QR login) to generate fresh cookies tied to your account, and then delete the included cookies file. 3) If the included cookies belong to you and you intentionally want to use them, rotate those session tokens/passwords after distributing code publicly. 4) Run the skill in an isolated environment (VM or container) if you must test it. 5) Review the code (browser.js, xhs-tools.js, ensure-service.js) for any additional endpoints or logging you don't expect; while no obvious exfiltration endpoints are present, the presence of embedded credentials is already a serious privacy/risk issue. If you don't control the account represented by the cookies, treat the package as unsafe and prefer a version that does not ship credentials.

Type: OpenClaw Skill Name: skill-xhs-mcp-service Version: 1.1.0 The skill bundle provides comprehensive XiaoHongShu automation but contains a highly irregular 'data/cookies.json' file pre-populated with active session tokens (including 'access-token' and 'web_session'), which is a significant security risk and irregular for a skill distribution. Additionally, 'scripts/browser.js' initializes Puppeteer with high-risk flags such as '--disable-web-security' and '--no-sandbox', and 'scripts/ensure-service.js' uses 'child_process.exec' for process management. While these elements are functionally related to the skill's purpose, the inclusion of hardcoded session data and the reduction of browser security boundaries are characteristic of high-risk or poorly secured automation tools.