fetching-blocked-urls

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [Data Exposure & Exfiltration] (LOW): The skill sends target URLs to https://r.jina.ai. This involves sharing potentially sensitive data (such as URLs with session tokens or internal addresses) with a third-party service not listed in the trusted sources.
  • [Indirect Prompt Injection] (LOW): The skill is designed to fetch and convert external web content into markdown for the agent to process. This creates a vulnerability surface where malicious instructions on external websites could influence agent behavior. No explicit boundary markers or sanitization steps are defined to mitigate this.
  • [Command Execution] (LOW): Employs a bash loop and curl to handle retries. While the commands are localized and hardcoded to the Jina service, it represents a pattern of script execution for network operations.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 01:00 PM