Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed academic paper restructuring skill with low system risk, though users should be aware it is Chinese-oriented and may produce a Word document at the end.
Install this if you want a structured Chinese academic-paper revision workflow. Before using it, confirm the target language, citation style, and whether you want a .docx file generated; for lightweight literature search or non-Chinese journals, this workflow may be more rigid than needed.
The trigger list includes generic phrases such as '帮我整理这篇论文' and '搜集文献' that could activate the skill for broad, everyday requests not specifically requiring this workflow. Over-broad activation increases the chance of unintended routing, causing the agent to apply a rigid multi-phase process, literature-search behavior, or file-generation flow when the user only wanted lightweight assistance.
The overview states the workflow operates under a fixed Chinese-language academic framing, and later formatting requirements specify Chinese fonts and conventions, but the skill does not offer the user a language or locale option. Under the policy, forcing a specific language/locale without opt-in is a natural-language policy violation unless clearly justified as region-specific.
The template hard-codes Chinese-language output and Chinese formatting conventions across the workflow without checking user preference or target journal requirements. While not a classic exploit, this is a policy and safety quality issue because it can override user intent, generate unsuitable outputs, and push locale-specific formatting assumptions into downstream file generation or publication workflows.
The workflow adds a seventh phase for Word document generation even though the skill metadata describes a six-phase paper reorganization process. This scope expansion can cause an agent to invoke file-generation behavior the user did not request and that the manifest did not disclose, increasing the chance of unintended tool use and output-side effects.
The instructions direct the agent to use an external docx-generation capability despite the skill’s stated purpose being paper restructuring and deepening. Unjustified tool invocation broadens the action surface from text transformation to artifact creation, which may trigger external capabilities, produce files containing hallucinated citations, or bypass user expectations about what the skill is allowed to do.
The skill culminates in generating a .docx deliverable and presenting it via a tool, but the top-level description does not clearly warn that it may create or modify a user-facing deliverable. This can undermine informed consent and surprise users with file-output behavior, especially if they expected only analysis or drafting assistance.
SQP-3 applies to all file types and covers language or locale policy violations. This markdown file presents all instructions exclusively in Chinese without any opt-in, alternative language option, or justification that the skill is intended only for a Chinese-language context.
No suspicious patterns detected.
