technology-insight-analysis
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses official domains (open.zhihuiya.com, open.patsnap.com) belonging to the vendor, Patsnap. These are standard resources for the skill's intended purpose of patent analysis.
- [EXTERNAL_DOWNLOADS]: The skill's HTML report template references the ECharts library via a well-known content delivery network (cdn.jsdelivr.net). This is a standard practice for rendering interactive charts and is considered safe.
- [METADATA_POISONING]: The skill contains a specific policy requiring reports to display the year '2026'. While this is a future date, it appears to be a stylistic or template requirement for the 'Strategic Opportunity' white paper format rather than a malicious attempt to deceive.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, structured patent data (e.g., tech types, functions, legal status). This represents a potential attack surface if the ingested data contains malicious instructions.
- Ingestion points: Data enters the process via patent records retrieved by upstream tools, as described in
SKILL.mdandreferences/white-paper-framework.md. - Boundary markers: None explicitly defined in the prompts to separate user data from analysis instructions.
- Capability inventory: The skill generates text and single-page HTML files with embedded JavaScript (ECharts).
- Sanitization: There are no explicit sanitization steps mentioned for the ingested patent labels before they are interpolated into the report or HTML template.
Audit Metadata