paper-style-toolkit
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The toolkit provides legitimate utility scripts and Jinja2 templates for academic document processing. All analyzed scripts (mineru_converter.py, style_extractor.py, etc.) perform local file operations without suspicious side effects or hidden remote execution. The dependencies (mineru, jinja2) are well-known and reputable.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function—analyzing academic papers—introduces an indirect prompt injection surface where malicious content within processed PDFs or Markdown files could theoretically influence downstream agents. This surface exists by design and is managed by structured analysis rather than raw text interpolation.
- Ingestion points: Ingestion occurs through
scripts/mineru_converter.py(PDF) andscripts/style_extractor.py(Markdown). - Boundary markers: The Jinja2 templates in
assets/do not currently implement explicit boundary markers or bypass warnings for analyzed content. - Capability inventory: Generated agent instructions include capabilities for file modification (
Write,Edit) and sub-agent invocation (Task). - Sanitization: The
style_extractor.pyutility sanitizes input by filtering it through specific Regular Expression patterns for statistical linguistic data rather than processing arbitrary commands.
Audit Metadata