linkfox-1688-search-by-image
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes and displays product metadata from the 1688 platform, which constitutes an untrusted data source and potential indirect prompt injection surface.\n
- Ingestion points: Product titles, seller identity badges, and descriptions retrieved from the /alibaba1688/imageSearch endpoint.\n
- Boundary markers: The skill instructs the agent to present data in structured tables, providing visual delimitation but lacking explicit instructions to ignore embedded commands.\n
- Capability inventory: The skill uses Python scripts with network access (urllib/requests) and file system write access within the local project directory.\n
- Sanitization: No explicit sanitization or filtering of external product content is implemented before display.
Audit Metadata