nature-paper-to-patent

Pass

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses subprocess.run to orchestrate internal utility scripts for rendering flowcharts, validating drafts, and generating Office documents. These operations are conducted using explicit script paths and specific binaries, avoiding high-risk shell environments.
  • [EXTERNAL_DOWNLOADS]: The skill interfaces with the official National Intellectual Property Administration (CNIPA) website (epub.cnipa.gov.cn) via Playwright. This functionality is used to retrieve patent abstracts and links for prior-art analysis, a core requirement of the drafting workflow.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted data from PDFs, Word files, and source code. This risk is managed through a grounded extraction process that maintains strict traceability between source text and generated patent claims.
  • [DYNAMIC_EXECUTION]: Internal scripts utilize importlib and shutil.which to dynamically locate and load the skill's own validation modules and external drawing tools like the Mermaid CLI. This allows for a modular architecture while remaining restricted to the skill's internal directory and the user's environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 12, 2026, 02:25 AM