firecrawl

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and scrapes content from arbitrary public URLs (e.g., commands like "firecrawl scrape https://example.com", "firecrawl search", "firecrawl crawl", and "firecrawl browser open ") and returns that untrusted, third‑party web content (in .firecrawl/ output files) which the agent is expected to read or process, enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's runtime "firecrawl scrape" and "firecrawl agent" commands fetch arbitrary external URLs (e.g., usage shown as --urls https://example.com) and feed the scraped content into the LLM/agent pipeline, meaning remote web content can directly influence agent prompts (prompt‐injection risk) and is relied on as a required runtime dependency.