firecrawl
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill implements a comprehensive interface for the Firecrawl web scraping service. All network operations are directed to the official Firecrawl API (
api.firecrawl.dev) and the DeepWiki service (deepwiki.com), both of which are legitimate and well-known entities in the AI research and scraping domain. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external web content into the agent's context, creating an indirect prompt injection surface. To mitigate this risk, the skill provides a
filter_web_results.pyscript for sanitization and explicitly instructs the agent to follow a 'search → filter → scrape → filter → reason' workflow to reduce the volume of untrusted data processed. - [COMMAND_EXECUTION]: The skill includes an 'interact' feature that allows for the execution of Node.js, Python, or Bash code. This code is executed within Firecrawl's remote, sandboxed browser environment as a standard service feature, providing safe interaction with dynamic web content without affecting the user's local system.
Audit Metadata