firecrawl-scrape
Warn
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from arbitrary external URLs, creating a vulnerability surface where malicious instructions in the scraped content could influence the agent's behavior.
- Ingestion points: Web content retrieved via
firecrawl scrapeandfirecrawl batch-scrapeinSKILL.md. - Boundary markers: Absent; there are no instructions to use delimiters or to ignore instructions embedded in the scraped data.
- Capability inventory: The agent can write to the local filesystem using the
-oflag and execute shell commands in a remote workspace using thefirecrawl/bashcapability. - Sanitization: Absent; the skill does not guide the agent to sanitize or validate the external content before processing it.
- [DYNAMIC_EXECUTION]: The skill utilizes a remote Bash capability (
firecrawl/bash) that executes commands provided as strings in the tool's options. - Evidence: Examples in
references/large-results.mdshow the construction of shell commands likejq,wc,head,sed, andgrepwhich are passed to thecommandfield at runtime. - [REMOTE_CODE_EXECUTION]: The
firecrawl/bashprovider enables the execution of code on remote workspaces managed by the vendor. While described as a restricted environment, it allows for remote execution of commands passed via the--optionsargument to thefirecrawl scrapecommand. - [COMMAND_EXECUTION]: The skill uses the
Bashtool to executefirecrawlandnpx firecrawl-cli. These commands involve arguments like--schema(JSON strings) and--query(natural language questions) that can be influenced by untrusted data or user input.
Audit Metadata