patent-drafting
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- SAFE (SAFE): The skill files (SKILL.md, patterns.md, sharp_edges.md, validations.md) contain only markdown-formatted guidance, legal principles, and regex patterns for detecting patent drafting errors.
- Indirect Prompt Injection (LOW): The skill is designed to ingest and analyze user-provided patent drafts. While it does not define specific boundary markers for untrusted input, it has no associated executable tools, shell access, or network capabilities, making the risk of exploitation negligible.
- Metadata Analysis (SAFE): The metadata and validation files use terms like 'Severity: critical' or 'Severity: high'. These are explicitly scoped to legal/patent rejection risks (e.g., 35 USC 112 rejections) and do not indicate security threats to the host system.
Audit Metadata