patent-panorama-insights-search
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill and its associated methodology files are consistent with the stated purpose of patent research and taxonomy construction. No malicious patterns such as obfuscation, persistence mechanisms, or unauthorized privilege escalation were identified.
- [INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted data from external patent databases, creating a theoretical attack surface for indirect prompt injection.
- Ingestion points: Patent data (titles, abstracts, IPC codes) retrieved through the Patsnap Open Platform (MCP) tools as described in
SKILL.md. - Boundary markers: None explicitly defined in the instructions to separate data from instructions.
- Capability inventory: File system write access limited to the session output directory (
@session/pps-output/) for generating.json,.csv, and.txtreports. - Sanitization: No specific sanitization methods for ingested patent text are mentioned, though the structured extraction tasks (Step 1-5) limit the impact of potential injections.
Audit Metadata