research

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

• Third-party content exposure detected (medium risk: 0.30). This skill’s runtime workflow explicitly uses Firecrawl to perform broad searches and deep scrapes of URLs from the open web (firecrawl_search -> firecrawl_scrape), which can include outsider-authored free-text content that the LLM will ingest.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly calls external MCPs (Context7 and Firecrawl) to fetch and scrape documentation at runtime — for example the firecrawl_map example uses "https://docs.example.com" — which would inject remote content into the agent's context and directly control its outputs.