Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed Xiaohongshu collection tool, but it enables broad copying of posts, comments, creator profile data, Feishu export, and recurring monitoring without enough privacy, retention, or opt-in controls.
Install only if you are comfortable with a skill that can scrape and preserve substantial Xiaohongshu content, creator profile data, and comments, and can export that data to Feishu or recurring notifications. Before using it, require explicit confirmation for each collection scope, Feishu export, and scheduled job; avoid storing full comments or personal/profile fields unless necessary; set retention limits; and make sure use complies with Xiaohongshu rules and applicable privacy or copyright obligations.
The workflow explicitly directs the agent to record all original note text and all original comments and then persist them to documents or Feishu tables. In context, this is more dangerous because the skill is built for systematic scraping and structured export, making large-scale collection, retention, and redistribution of third-party content a core behavior rather than an incidental side effect.
The README promotes data collection, creator profiling, scheduled tracking, and Feishu output, but does not provide clear privacy notices, consent requirements, retention expectations, or warnings about sending collected data to external services. In this context, the combination of profiling and automated export materially raises the risk of privacy violations, unauthorized monitoring, and accidental exfiltration of personal or platform-derived data.
The invocation examples are very broad and do not define clear trigger boundaries, which increases the chance the skill will activate for ambiguous user requests involving scraping, profiling, scheduled monitoring, or external export. In an agent environment, overbroad activation can cause unintended collection or transmission of data beyond what the user specifically intended.
The skill advertises Feishu table output but does not clearly warn that collected data may be transmitted to an external third-party service. Because the skill gathers note content, comments, and creator profile metrics, silent export could expose scraped content or personal/profile data beyond the immediate session, creating privacy, compliance, and data handling risks.
The trigger examples are broad, generic user requests such as collecting information, analyzing topics, tracking creators, and exporting results. This can cause the skill to activate for common requests without clear user intent boundaries, increasing the chance of unnecessary web scraping, profile collection, or downstream external export actions being performed in contexts where the user did not explicitly request them.
The instruction to preserve nearly all original content and all comments encourages bulk verbatim reproduction of scraped user-generated text, which can include personal, sensitive, or copyrighted material. At scale, this meaningfully increases the chance of privacy leakage and inappropriate redistribution of third-party content.
The skill instructs saving collected platform data to local documents and Feishu tables without any explicit user-facing disclosure about what data will be stored, where it will go, or whether third-party systems will receive it. This creates a real privacy and data-handling risk, especially because the collected content includes creator data, links, and potentially comment text at scale.
The scheduled-tracking feature enables continuous collection and outbound delivery via group messages or email, yet the document lacks a clear warning about ongoing monitoring, repeated data transfer, and accumulation over time. Persistent automation materially increases privacy, compliance, and misuse risk because data can be collected and redistributed long after the initial request.
该技能在“指令索引”中使用了“帮我分析xxx话题的爆款笔记”“把结果保存到飞书表格”“设置一个定时任务跟踪xxx”等自然语言示例，但没有说明这些是否为完整触发短语、是否需要特定上下文，或哪些相似表达不应触发。由于缺少边界条件和排除示例，这些常见表达可能在普通对话中与技能调用发生冲突。
The sample output normalizes including full original text and verbatim comments in reports, which encourages downstream sharing of scraped content beyond the source platform. Even as a template, it institutionalizes unsafe data handling and makes privacy leakage more likely during normal use.
“使用场景”表中的表达如“采集这个博主的信息”“分析这个博主的画像”“每天早上汇报热点”都属于高频自然语言，文档未限制它们仅适用于小红书领域，也未给出必要参数或不适用情形。对于基于自然语言匹配的技能系统，这种描述会增加跨场景误调用风险。
All user-facing instructions, examples, and labels in this file are Chinese-only. Under the policy, forcing a specific language without user opt-in can be a natural-language policy concern unless the locale limitation is explicitly documented and justified.
No suspicious patterns detected.
