Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed WeChat article-writing assistant that stores writing profiles locally and only archives to WeCom Docs after user confirmation.
Before installing, be aware that the skill keeps your account profile, style notes, active account, and article index in the workspace. Only confirm WeCom Docs archiving for content you are comfortable sending to that service, and review or delete the local reference files if they contain sensitive editorial or business information.
The high-level description emphasizes functionality but does not clearly warn users up front that profile data is stored locally in the workspace and that articles may be archived to WeCom Docs. This can undermine informed consent, especially because the skill handles persistent writing profiles and potentially sensitive unpublished content across multiple accounts.
The documented activation phrases are broad natural-language requests such as asking what to write or asking for article help. In an agent environment, overly generic triggers can cause accidental invocation during ordinary conversation, leading the skill to access or modify local workspace state, switch accounts, or begin content archival flows without clear user intent.
The trigger table includes ambiguous phrases like '今天写什么' and '帮我写公众号' without any explicit opt-in boundary. Because the skill manages persistent profiles and archives content, ambiguous activation increases the chance of unintended execution and privacy-affecting actions in response to casual user text.
The trigger list includes broad, everyday phrases such as '写文章' and '这篇文章写得好', which can match routine conversation and cause the skill to activate unexpectedly. Because the skill reads and writes workspace files and can invoke external capabilities like web_search and wecom-doc, unintended invocation can expose or modify user data and lead to actions the user did not explicitly intend.
The skill behavior stores user-provided account profiles, style preferences, article indexes, and potentially article metadata in workspace files, but the description does not clearly warn users up front about this persistence. This weakens informed consent and increases the risk that users share sensitive business, editorial, or personal information without realizing it will be retained and reused across sessions.
The example trigger phrases "更新公众号风格" and especially "这篇文章写得好" are natural, everyday expressions without clear constraints on when they should invoke the skill. The file does not provide scope limits, explicit activation rules, or negative examples to distinguish ordinary conversation from intentional invocation.
The file is written as a universal writing standard and all directives are framed exclusively in Chinese, beginning with the title '通用写作规范'. Because it is described as applying broadly ('适用于所有公众号方向'), this can constitute a language-policy constraint without an explicit user choice or justification.
The file’s user-facing content is entirely in Chinese, including the title and maintenance notes, with no indication that language choice is optional or that the skill is intentionally limited to a Chinese-language context. Under the stated policy, forcing a specific language without user opt-in can be a natural-language policy violation.
