evaluate-patent-transfer-ip
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 patent data and technical disclosures retrieved via MCP connectors or web research. This untrusted content creates a surface for indirect prompt injection, where malicious instructions embedded in patent abstracts or news could attempt to manipulate the agent's evaluation, ranking, or the content of the generated brief.
- Ingestion points: Patent metadata (bibliography, claims, descriptions) and web-based signals retrieved from Patsnap MCP services and external research (SKILL.md).
- Boundary markers: Not specified for the internal reasoning phase; however, the skill mandates strict boundaries for the final report output.
- Capability inventory: Data synthesis, counterparty ranking, and self-contained HTML report generation.
- Sanitization: Mandatory escaping of all retrieved or user-supplied content and an explicit prohibition of remote scripts, CDNs, or executable content in the generated HTML brief.
Audit Metadata