scheduled-task

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly instructs building prompts that tell Claude to "搜索并整理今天最重要的 AI 领域新闻" / "搜索今天的科技新闻热点" and similar (see the example prompts in Principles and Examples), and it states that the prompt is the sole runtime instruction and that "all tool calls auto-approved", so the agent is expected to fetch and interpret open/public web content (news/weather) which could contain untrusted user-generated instructions and thus enable indirect prompt injection.