feasibility-review

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The skill strongly steers the agent to write and execute a local single-shot Python script, forbid stepwise/chat outputs, and persist results to session/site files (including a web platform data folder), which is suspicious steering that reduces transparency though not explicitly malicious.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). 技能运行时会读取用户上传的可行性研究报告/项目指南进行文本解析，并把从报告中提取的路线/创新点关键词用于专利与论文检索（patent.search/mcp_patent-search__patsnap_search sources=paper）以及用 web-search 搜索政策，从而导致外部提交的报告自由文本进入LLM/流程上下文用于后续生成。