Security audit
Security checks for vulnerabilities and agentic risk
This Chinese report-writing skill is coherent and purpose-aligned, but users should be mindful that it can use recent conversation context and optionally create Feishu documents.
Install this only if you want a Chinese-language Feishu reporting helper. Before enabling Feishu document creation, check the create_doc setting and destination/account expectations, and avoid asking it to summarize confidential conversations unless your organization allows that content to be written to Feishu.
The trigger phrases are broad, common workplace expressions such as '月报', '项目报告', and '会议纪要', which can plausibly appear in ordinary conversation without the user intending to invoke the skill. Because the skill can read recent conversation history and create or update Feishu cloud documents, accidental activation could cause unintended collection of context and external side effects.
The skill states that it reads recent conversation history to extract work content and creates or writes Feishu cloud documents, but it does not clearly warn the user or obtain explicit consent for those sensitive operations. This creates privacy and data leakage risk because internal discussions, confidential project details, or personal information could be silently aggregated and exported to an external document system.
The handler can automatically send AI-generated report content to Feishu using a user identifier without any visible confirmation, warning, or consent check. In a reporting skill, the content may include sensitive work details, so silent exfiltration to an external collaboration platform creates a privacy and data-governance risk even if the destination is legitimate.
The manifest uses only Chinese natural-language fields for the display name, description, and tags, which suggests the skill is presented in a fixed language without any stated user opt-in or region-specific justification. Under the policy rules, locale constraints should either offer user choice or be clearly documented as intentional and scoped.
Comments, report type detection, text extraction, and the returned success message all assume Chinese input/output behavior. This forces a specific language/locale without offering user opt-in or documenting a justified region-specific limitation.
The inline comment at L10 says the code is generating a report, but the subsequent logic conditionally creates a Feishu document and returns a document reference. This is not just omitted detail in a helper; the nearby comments/documentation frame the operation as report generation while the code performs an additional external document-creation action.
No suspicious patterns detected.
