doc-fetcher

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

• Third-party content exposure detected (high risk: 0.95). SKILL.md describes runtime fetching of external documentation pages via browser_links(base_url) and browser_markdown(page_url), which ingests outsider-authored free text from public web content into local markdown for later LLM/RAG context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls browser_links/browser_markdown at runtime to fetch external documentation (e.g., "https://hono.dev", "https://developers.cloudflare.com/workers/", "https://docs.astro.build/") and injects that content as context for RAG, meaning remote pages directly influence prompts.