firecrawl
Pass
Audited by Gen Agent Trust Hub on Apr 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
Bashtool to executefirecrawlandnpx firecrawlcommands for web searching, scraping, and browser automation as described inSKILL.md. - [EXTERNAL_DOWNLOADS]: The skill documentation in
rules/install.mdandrules/security.mdguides the installation of thefirecrawl-clipackage from the npm registry. - [REMOTE_CODE_EXECUTION]: The
firecrawl browsercommand features anevalcapability which allows for the execution of arbitrary JavaScript code within a remote browser environment. - [PROMPT_INJECTION]: The skill processes data from external websites which presents an indirect prompt injection surface. The
rules/security.mdfile defines a security model to handle this untrusted data. - Ingestion points: Untrusted web content is retrieved via
scrape,search, andcrawlcommands. - Boundary markers: Use of the
-oflag is encouraged to isolate fetched content within the.firecrawl/directory. - Capability inventory: The skill has the ability to write to the local filesystem and perform network operations through the CLI tool.
- Sanitization: The documentation recommends quoting URLs and using incremental reading tools like
greporheadto limit exposure to potentially malicious content embedded in web pages.
Audit Metadata