email-manager

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

• Third-party content exposure detected (high risk: 0.95). 在运行时，scripts/check_email.py 通过 email_client.py 的 IMAP fetch_unread() 读取“未读邮件”的 Subject/From/Body（外部邮件正文由邮件发送方撰写），随后 generate_ai_summary() 将这些字段拼入摘要文本并写入 cache/last_summary.txt，该摘要会进入代理/LLM上下文用于通知与回复流程。

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

• Potentially malicious external URL detected (high risk: 0.80). The skill connects at runtime to IMAP servers (e.g., imap.gmail.com, imap.qq.com, outlook.office365.com, imap.163.com, imap.126.com) and fetches email bodies that are then injected into the AI summary/reply generation, so remote email content can directly control model prompts.