Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent Chinese-platform writing assistant that fetches Zhihu trend data and drafts content, with limited but real network and local-save side effects users should understand.
Install only if you want a Chinese Zhihu/Yanyan content workflow. Expect it to fetch trend data from tophub.today, produce public-facing draft content, and potentially save drafts under ~/Documents/notes; confirm before creating any scheduled monitoring task or publishing/posting content.
This markdown file presents the skill examples, instructions, and usage entirely in Chinese, and does not indicate that the user can choose another language. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is explicitly justified.
All user-facing instructions and examples are presented only in Chinese, and the skill does not indicate that language choice is optional or that the Chinese-only scope is a documented requirement. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation.
The invocation examples are broad natural-language phrases that could accidentally trigger the skill in normal conversation without a clear activation boundary. In an agent setting, this raises prompt-routing and unintended tool-use risk, especially because the skill may fetch external trending data and generate publishable content from a casual user utterance.
整份技能文档从标题、触发语到输出模板均默认且固定为中文，没有提供用户可选择的语言或本地化选项。根据语言/地区政策，若技能强制单一语言而无用户选择或充分说明，属于自然语言策略问题。
The skill instructs the agent to fetch data from an external site and later save outputs locally, but it does not clearly disclose these side effects to the user at the point of use. Hidden network access and file writes weaken informed consent and can lead to privacy, compliance, or trust issues if triggered unexpectedly.
The trigger phrases are broad natural-language requests that can overlap with ordinary conversation, making accidental activation plausible. In a skill that performs external fetching and content-generation workflows, unintended triggering can cause the agent to take actions the user did not clearly authorize.
The short-story writing triggers are open-ended and context-light, so loosely related user requests could invoke a large drafting workflow without clear consent. Because the skill can generate substantial content and later store outputs, ambiguous activation increases the risk of unwanted processing and unexpected downstream actions.
The dual-track workflow trigger is semantically vague, so the agent may infer multiple operations from a brief phrase and execute more than the user intended. This is especially risky because the workflow chains topic selection, analysis, writing, and possible promotion/storage behavior in one path.
The README states that the skill automatically retrieves Zhihu trending data from an external source (tophub.today) but does not warn users that outside data access will occur. This can undermine informed consent, surprise users in restricted environments, and increase trust in unverified third-party content used to drive writing recommendations.
No suspicious patterns detected.
