Security audit
Security checks for vulnerabilities and agentic risk
This skill clearly focuses on saving paper-analysis methods into a local database and does not show hidden execution, networking, credential use, or destructive behavior.
Install this if you want research-method notes from paper analysis saved persistently in a local SQLite database. Be aware that saved analysis can remain on disk at the disclosed database path, so avoid using it for confidential, proprietary, or unpublished material unless that local retention is acceptable.
The skill is user-invocable and its activation phrases are common conversational requests like '存入数据库' and '保存到知识库'. Without tighter intent boundaries or confirmation, the agent may invoke persistent storage during ordinary dialogue or when a user did not fully understand that local data would be written, leading to unintended data retention.
The skill performs writes to a local SQLite database at a fixed filesystem path but does not clearly warn users that information will be stored persistently. In this context, extracted paper analysis may include proprietary, unpublished, or sensitive notes, so silent persistence increases privacy, retention, and compliance risk.
The trigger list contains only positive examples and no scope limits, exclusions, or disambiguation rules. This increases the chance of accidental activation from nearby conversational text, quoted examples, or meta-discussion about saving methods, which can cause unintended database writes and persistence of possibly sensitive research content.
The manifest description and trigger phrases are entirely in Chinese, implying a fixed language expectation for invocation and use. The file does not indicate that this is optional, user-selectable, or intentionally limited to a Chinese-language context.
No suspicious patterns detected.
