Security audit
Security checks for vulnerabilities and agentic risk
This skill is a narrow OCR and keyword-marking helper with some overclaimed output wording, but no hidden, persistent, destructive, or credential-seeking behavior was found.
Install only if you want a Chinese-focused OCR and keyword-highlighting helper. Expect Markdown/text output rather than an actual generated Word document, and review or redact sensitive images before asking the agent to process them.
声明描述的是一个端到端的“图片 OCR → 关键词标注 → Word 文档生成”能力。但实际代码只接收纯文本文件作为输入，对其中关键词做正则替换加粗，并输出文本文件。代码中没有任何图像读取、OCR库调用、上传处理、批量图片处理或 doc/docx 生成逻辑。虽然“关键词标注加粗”这一子功能与声明部分一致，但核心输入类型（图片 vs 纯文本）和核心输出能力（Word 文档生成 vs 文本保存+转换提示）均存在实质差异，因此属于明显不匹配。
Without declared permissions the skill's intent is opaque and cannot be validated.
The docstring states '生成可用于 Word 的格式', which implies Word-oriented output, while the code at bolding time inserts Markdown syntax and later instructs the user to manually convert it. This is an active documentation-to-code mismatch rather than a minor omission, because the stated output format is materially different from the actual one.
The manifest describes a pipeline that accepts images, performs OCR, and outputs a continuous Word document. In the code, there is no image handling, no OCR logic, and no Word document creation; it only processes a UTF-8 text file and wraps keyword matches with Markdown '**' markers before saving text.
The skill processes uploaded images and performs OCR on potentially sensitive text, but it gives no warning that image contents may contain personal, academic, financial, or confidential information. This can lead users to upload sensitive documents without informed consent or redaction, increasing privacy and data-handling risk.
This plain-text output presents all instructional and summary content in Chinese only. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation, and there is no indication here of user choice or a justified locale constraint.
The natural-language strings in the module docstring and argument help are entirely Chinese, which imposes a specific language on users. The file does not offer an opt-in language choice or explain that the tool is intentionally limited to a Chinese-speaking or region-specific context.
The role definition states the assistant is focused on extracting Chinese text, which imposes a language constraint in natural-language instructions. The file does not offer opt-in language selection or explain why Chinese-only handling is required for this skill.
