exampass

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

• Third-party content exposure detected (high risk: 0.85). 默认/graph流程会把“从用户未选择引入的外部课程材料”提取出的自由文本（PPT/DOCX/PDF内容、OCR结果、以及 slide_renderer 渲染时附带的原文文字）作为可读 prose 注入到子 Agent 的 LLM prompt（如 knowledge_graph.py / knowledge_analyzer.py / notes/item prompts），属于 outsider-authored content（课程资料作者非操作用户）。

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). High-confidence: the included update flow (/exampass update) runs git fetch/git pull from the remote origin (the repo cloned in README: https://github.com/WUBING2023/ExamPass-Assistant.git) and then runs pip install -r requirements.txt at runtime, which fetches remote code/updates that can change agent prompts and execute code.