geolocation
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs legitimate geolocation analysis using visual cues and web verification without any signs of malicious intent or unauthorized actions.
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external data sources as part of its primary geolocation function.
- Ingestion points: Image files processed by
scripts/examine_region.pyand web search content viaWebSearchandWebFetchtools. - Boundary markers: The instructions include privacy-focused constraints, such as prohibiting the identification of individuals or residential addresses.
- Capability inventory: The skill uses local image manipulation (Pillow) and standard web-access tools.
- Sanitization: The skill relies on iterative reasoning and multiple independent confirmations to validate findings from untrusted sources.
Audit Metadata