patent-panorama-insights-tag
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted technical text from external files like 'tagged_pool.csv' and 'candidate_pool.csv'. This creates a potential surface for indirect prompt injection within the patent technical descriptions.
- Ingestion points: 'SKILL.md' defines these CSV files as mandatory inputs.
- Boundary markers: The skill includes an explicit instruction to minimize context contamination by avoiding reading full rows into the LLM context during data export ('绝不把行读入上下文').
- Capability inventory: The skill calls the '/7cc6ae/mcp' tool and performs multiple file-write operations to the '@session/pps-output/' directory.
- Sanitization: No explicit content filtering is mentioned, but the workflow relies on platform-provided tools for analysis.
- [EXTERNAL_DOWNLOADS]: The README provides an installation command that fetches resources from GitHub.
- Evidence: 'npx skills add patsnap/skills/tree/main/open-platform/patent-panorama-insights-tag'.
- Context: This reference points to the official Patsnap GitHub organization and is part of the standard deployment process for skills in this environment.
Audit Metadata