claims-drafting
Pass
Audited by Gen Agent Trust Hub on May 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured instructions for patent claim drafting, adhering to international standards (e.g., USPTO, EPO, CNIPA guidelines). It includes clear logic for different patent types and jurisdictions.
- [COMMAND_EXECUTION]: While the frontmatter allows the 'Bash' tool, the instructions themselves do not contain any arbitrary command execution or system-level operations. The tool is likely used for file management within the agent's workspace.
- [EXTERNAL_DOWNLOADS]: The skill references local files (e.g., '../shared-references/patent-writing-principles.md') for guidance rather than external URLs. No suspicious downloads from untrusted sources were detected.
- [REMOTE_CODE_EXECUTION]: The skill uses 'mcp__codex__codex' for a peer-review step (Step 5). This is a call to a structured model for quality control of the drafted text, not execution of untrusted code. The model parameters are set to high-reasoning mode for better text analysis.
- [PROMPT_INJECTION]: No attempts to bypass safety filters or override agent behavior were found. The use of 'IMPORTANT' and 'CRITICAL' markers is limited to instructional formatting and legal formatting requirements (e.g., claim numbering rules).
Audit Metadata