book-illustrator

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

• Third-party content exposure detected (high risk: 0.95). 运行时会从“用户提供的 URL”下载外部章节 Markdown（OUTSIDER：公共 web 内容/任意 URL），并将其前后 1-3 段原文整段内联进 subagent brief，从而进入 LLM 上下文。

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly supports downloading a chapter Markdown at runtime (e.g., https://example.com/book/chapter-03.md) and then inlines that fetched text into subagent briefs that form the prompts, so remote content fetched during execution directly controls the agent prompts.