Security audit
Security checks for vulnerabilities and agentic risk
This skill coherently helps analyze user-provided PDFs by rendering pages to images for AI vision, with privacy and dependency caveats but no hidden or destructive behavior found.
Before installing, use this only with PDFs you are comfortable processing in your current AI environment, avoid archiving sensitive document contents to memory unless you explicitly want that, and install dependencies in an isolated environment because the package versions are not pinned.
README.md:13Unpinned and Unnecessary Third-Party Dependencies
The code matches only one declared sub-capability: rendering PDF pages to high-resolution PNG images. The broader declared purpose emphasizes visual understanding of PDF content via an AI model, multilingual analysis, summarization, and structured interpretation of elements like tables and charts. None of that behavior exists in the provided code chunk. The code is instead a standalone extraction utility plus basic metadata inspection. This is a material description-behavior mismatch because the declared primary purpose is end-to-end PDF visual comprehension, while the actual code only implements preprocessing/image extraction.
The README explicitly describes converting user PDFs into images and sending them to an AI vision model, but it does not disclose that sensitive document contents may leave the local processing boundary for third-party analysis. Users may unknowingly submit confidential, regulated, or personal data, creating privacy, compliance, and data-handling risks.
The activation description is broad enough that the skill may trigger on many ordinary PDF-related requests, causing unintended handling of user documents and routing content to a vision-analysis pipeline. In context, this increases the chance of over-collection and unnecessary exposure of potentially sensitive PDF contents, especially if users did not explicitly request image rendering or AI vision processing.
The skill explicitly mentions optional archiving to the memory/ directory without any visible consent, retention, or sensitivity warning. Because PDFs often contain confidential or personal data, silent or loosely prompted persistence can create privacy, compliance, and data-retention risks if content is stored beyond the immediate task.
Natural-language policy violations apply to all file types, including markdown. The file forces a specific language for instructions and usage details without user opt-in or an explicit note that the skill is region- or language-specific, which can violate language/locale policy expectations.
The manifest description is written entirely in Chinese, which imposes a language choice in user-facing natural language without indicating any option for alternative locales. The policy specifically calls for flagging language or locale constraints when they are forced without user opt-in or justification.
No suspicious patterns detected.