Security audit
Security checks for vulnerabilities and agentic risk
This skill appears to be a straightforward WeChat article archiving helper, but it will write persistent knowledge and memory files.
Install this only if you want a write-enabled workflow for archiving WeChat articles. Before running it, provide the exact article source and confirm the target knowledge-base files and MEMORY.md changes so unrelated article-reading or summarization requests do not create durable entries by accident.
The skill can be invoked for broadly defined requests such as obtaining article content or updating process documentation, without clear scope boundaries or confirmation gates. In practice, this increases the chance the agent will trigger the workflow unnecessarily and perform browsing plus local file modifications in contexts the user did not explicitly authorize.
The workflow instructs the agent to update archive files, process documents, and long-term memory, but it does not clearly warn the user that multiple local files will be modified. This can lead to silent persistence of data and unintended changes to knowledge or memory stores, which is risky in an agent environment because users may not expect durable writes.
The eval prompts are written to trigger on very generic requests about archiving WeChat articles and repeatedly expect the system to update archive files, procedural documents, and long-term memory without explicit scoping, authorization, or safety boundaries. This can cause over-broad skill activation and unintended persistent writes based on loosely matched user requests, increasing the risk of inappropriate data ingestion or autonomous state changes.
The skill title, usage conditions, steps, and task record are entirely in Chinese, which effectively fixes the operating language without user opt-in. Under the language/locale policy rule, a skill should either offer a language choice or clearly justify a locale-specific restriction.
No suspicious patterns detected.
