us-provisional-patent
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill was reviewed for prompt injection and concealment patterns. The flagged 'concealment' heuristic appears to be a false positive triggered by safety instructions that explicitly limit the agent's persona (e.g., 'Do not present yourself as a lawyer' and 'Do not claim that the skill filed... anything'). These are transparency measures rather than attempts to hide malicious behavior. The skill proactively mandates a 'Startup Safety Notice' to be shown to the user, enhancing transparency.
- [EXTERNAL_DOWNLOADS]: The skill references the official USPTO website (uspto.gov) for fees and filing. These are links to well-known government services and do not involve automated data fetching or execution. No automated external downloads are present.
- [DATA_EXFILTRATION]: No network operations, file system access, or credential harvesting patterns were detected. The skill only processes text provided by the user within the conversation context.
- [NO_CODE]: This skill consists entirely of markdown-based instructions and guidelines. There are no associated scripts, executables, or package dependencies, which significantly minimizes its attack surface.
Audit Metadata