geoflow-content-automation

Warn

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill documents default administrative credentials for the system, which poses a security risk if not changed immediately after deployment.
  • Evidence: Default username 'admin' and password 'admin888' are provided for the admin panel in SKILL.md.
  • [EXTERNAL_DOWNLOADS]: The skill downloads and runs software from an external repository not associated with the primary vendor.
  • Evidence: The installation instructions use 'git clone' to fetch the GEOFlow source code from a third-party GitHub repository (yaojingang/GEOFlow).
  • [INDIRECT_PROMPT_INJECTION]: The system processes content from external title and knowledge libraries which are interpolated directly into prompts, creating a surface for manipulation of the AI's behavior.
  • Ingestion points: The system pulls data from a 'Title Library' and 'Knowledge Base' defined in SKILL.md.
  • Boundary markers: No explicit boundary markers or delimiters are shown in the prompt template logic.
  • Capability inventory: The worker processes jobs that call AI APIs to generate articles.
  • Sanitization: Content is interpolated into templates using simple string replacement without validation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 12, 2026, 11:44 PM
Security Audit — agent-trust-hub — geoflow-content-automation