patent-panorama-insights-report
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from
tagged_pool.csvandpatent_packages.csvwhich could contain untrusted user-supplied content. - Ingestion points:
tagged_pool.csv(SaaS data export) andpatent_packages.csv(manual tags). - Boundary markers: The instructions provide a strong mitigation by explicitly forbidding the agent from reading large CSV data line-by-line into the prompt context, requiring script-based aggregation instead.
- Capability inventory: The skill is authorized to write local files (
report.html,report_manifest.json). - Sanitization: The instructions emphasize using the vendor's MCP (Model Context Protocol) tools and pre-aggregated data results rather than raw user input strings where possible.
- [EXTERNAL_DOWNLOADS]: The skill references Model Context Protocol (MCP) endpoints (e.g.,
/59100a,/3fd502) which are official services provided by the developer's platform (Patsnap/Zhihuiya) for fetching patent analytics data.
Audit Metadata