inventor-resignation-monitor
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from external patent databases.
- Ingestion points: External patent data (titles, abstracts, and inventor names) is retrieved via the
mcp_patent-search__patsnap_searchtool as described in thereferences/workflow.mdfile. - Boundary markers: The instructions and prompt generation logic in
scripts/run_monitor.pydo not implement clear delimiters or negative constraints to prevent the agent from following instructions potentially embedded in patent metadata. - Capability inventory: The skill includes the capability to write HTML reports to the local file system using the
scripts/run_monitor.pyscript. - Sanitization: The report generation process performs raw string interpolation of patent data into HTML templates without sanitization or escaping, creating a surface for both injection into the agent's logic and potentially cross-site scripting (XSS) within the generated report.
- [EXTERNAL_DOWNLOADS]: The skill references an external configuration portal at
open.zhihuiya.com. This domain is the official service portal for the skill author (Patsnap) and is used for necessary MCP tool configuration and authorization.
Audit Metadata