Security audit
Security checks for vulnerabilities and agentic risk
This is a Chinese text-editing skill for making copy sound less AI-generated, with no evidence of hidden access, persistence, exfiltration, or destructive behavior.
Install this if you want Chinese-language copy-editing help to make text sound more natural. Be aware that it is tailored to Chinese phrasing and may activate on broad requests like asking text to sound more natural, but it does not request sensitive access or make system changes.
描述包含两类核心能力：检测和修复优化。实际脚本确实实现了‘检测AI味’这一部分，通过关键词/模式匹配识别套路开头、连接词、过度礼貌、模板结尾、书面词、结构词并累计分数，再输出诊断结果和简短建议。但代码没有任何对原文进行改写、润色、口语化重写或自动修复的逻辑，因此其主要行为只覆盖了声明的一半。未发现额外的越权行为、资源访问或无关触发器问题；不匹配点在于描述夸大了能力范围。
Overly broad trigger phrases can cause the skill to activate unintentionally in unrelated conversations, which increases the chance of prompt injection surface expansion and unexpected behavior. In this case the skill is low-risk because it only rewrites tone/style, but accidental invocation could still interfere with normal assistant behavior or cause inappropriate style changes.
This markdown file presents all examples and guidance exclusively in Chinese, and there is no indication that users can opt into another language or that the skill is intentionally limited to a Chinese-language audience. The policy requires flagging language or locale constraints when a skill forces a specific language without user opt-in.
The script’s comments, regex patterns, and user-facing output are entirely in Chinese, and its detection logic is specifically tailored to Chinese phrases. There is no indication that users can opt into another language or that the Chinese-only behavior is a documented regional constraint, which creates a language/locale policy concern.
No suspicious patterns detected.
