competitor-profiling

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

• Third-party content exposure detected (medium risk: 0.30). Outsider-authored free text is ingested because the workflow takes user-provided competitor URLs and uses Firecrawl Map/Scrape (and optionally Search) to fetch arbitrary third-party webpage/review content, then the LLM parses that scraped markdown into the competitor profiles.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly scrapes and ingests user-provided competitor URLs at runtime (see firecrawl_map / firecrawl_scrape) which will inject external page content into the agent context — examples in the evals include https://competitor1.com, https://competitor2.com, and https://competitor3.com.