deep-research

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

• Third-party content exposure detected (high risk: 0.95). The required workflow uses firecrawl/Exa to fetch and scrape public web pages at runtime (e.g., firecrawl_search/firecrawl_scrape and web_search_exa/crawling_exa), ingesting outsider-authored free text from those URLs into the LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches arbitrary external pages at runtime using calls like firecrawl_scrape(url: "") and crawling_exa(url: "", tokensNum: 5000), which will be injected into the agent context and can therefore directly control prompts or carry malicious instructions.