cn-biopharma-patent-drafting
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute local Python and Node.js scripts (AI60, AI32, AI37) located in '$USERPROFILE/.workbuddy/api-wrappers/'. These are documented vendor-provided CLI wrappers for document processing and generation, using specific environment variables or CLI arguments for authentication as per vendor guidelines.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided materials including technical descriptions, PDF documents, and images.
- Ingestion points: User-supplied files and text inputs processed in SKILL.md.
- Boundary markers: The instructions mandate a 'Source Map' to explicitly trace every technical feature back to specific user inputs or API outputs.
- Capability inventory: The agent can execute local CLI tools via Bash and perform extensive searches via biological and chemical MCP tools.
- Sanitization: The skill includes strict instructions to avoid fabricating technical features and to flag missing information, reducing the risk of malicious instructions in the source data influencing the output without validation.
Audit Metadata