小红书达人笔记质检 for bbt

对小红书达人笔记做品牌词、必带话题、@官号、调性、链接与广告法/竞品合规的自动化质检， 输出结构化报告；品牌名正确/错误写法分表（assets/brand-names-correct.txt / assets/brand-names-wrong.txt）， 并支持在 skill 目录内更新各词表文件。在用户粘贴小...

This skill is coherent with its goal: it reads local wordlists, fetches content from user-provided Xiaohongshu URLs, and runs checks described in SKILL.md. Things to consider before enabling: (1) The agent will make HTTP GET requests to URLs provided by users — those endpoints will see requests from the runtime environment. (2) The skill allows whole-file replacements of files inside its own skill directory when the user asks to '更新词表' — only allow this if you trust the agent/user input and prefer edits to be limited to non-executable asset files (txt/md). (3) The official-account validation logic is documented in a markdown file rather than a structured list; verify this meets your exact matching needs. If you want to tighten safety, require manual confirmation before the agent writes skill files and run the fetch scripts in a network-restricted sandbox.

Type: OpenClaw Skill Name: xhs-content-qc Version: 1.0.2 The skill bundle facilitates automated quality control for Xiaohongshu content by fetching external URLs and performing text analysis against local word lists. It includes a shell script (`fetch_xhs_content.sh`) and a Python script (`fetch_xhs_content.py`) that use `curl` to retrieve web content. A significant vulnerability exists in `SKILL.md`, which instructs the AI agent to execute these scripts using user-provided URLs, potentially leading to command injection if the input is not sanitized. Furthermore, the 'Word List Update' feature grants the agent the capability to overwrite local files in the `assets/` and `references/` directories. While these features align with the stated purpose of content moderation and configuration management, the combination of external network requests, shell execution, and file-writing capabilities presents a high-risk surface without evidence of direct malicious intent.