docx 审计报告.

警告

• SKILL.md does not mention that extracted Word document content should be treated as untrusted data, which is relevant for agents that might interpret document text as instructions.
• SKILL.md documents paragraph extraction but does not note that tables, headers, footers, and text boxes are ignored.
• No guidance on handling unusually large or malformed .docx files; python-docx parsing can consume significant memory if given a crafted or zip-bomb-like document.
• Quality score needs review

方法

本报告综合公开元数据、AI 审查输出、仓库活跃度、安装就绪度、OpenAgentSkill 事件、质量评分、信任检查和 Agent 安全门槛；它不是完整的源代码安全审计。