ato-rate-scraping

Warn

Audited by Gen Agent Trust Hub on Feb 14, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): Employs curl for network requests. Shell command usage increases the attack surface for the agent.
  • EXTERNAL_DOWNLOADS (MEDIUM): Retrieves data from r.jina.ai and Australian government websites. Content from these external sources is untrusted.
  • PROMPT_INJECTION (MEDIUM): Susceptible to indirect prompt injection via scraped web content.
  • Ingestion points: ATO websites via Jina Reader defined in Rate Source Registry.
  • Boundary markers: Absent.
  • Capability inventory: curl (network) and cache-manager.ts (file write).
  • Sanitization: Numeric range validation present, but natural language filtering is absent.
  • CREDENTIALS_UNSAFE (LOW): References $JINA_API_KEY. While not a hardcoded secret, it requires external credential handling and introduces a dependency on sensitive tokens.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 14, 2026, 03:58 PM