gzh-explosive-content-detector

Warn

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The script scripts/fetch_gzh_trends.py communicates with https://onetotenvip.com to retrieve trending content. This domain is not an established or verified service provider.
  • [DATA_EXFILTRATION]: User-provided search keywords are sent to an unverified third-party server (onetotenvip.com). Furthermore, the custom network implementation in fetch_via_no_sni explicitly disables SSL certificate validation and hostname verification, rendering the connection insecure and susceptible to Man-in-the-Middle (MITM) attacks.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting article summaries and titles from an external API.
  • Ingestion points: Article data retrieved from onetotenvip.com and processed by the agent.
  • Boundary markers: Instructions lack explicit delimiters or warnings to ignore instructions embedded within the fetched data.
  • Capability inventory: The skill allows writing search results to local HTML and JSON files.
  • Sanitization: While the script escapes HTML for the final report, it does not sanitize or filter external content before the agent analyzes it to generate recommendation reasons.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 22, 2026, 03:50 PM
Security Audit — agent-trust-hub — gzh-explosive-content-detector