krt-document-forge

Pass

Audited by Gen Agent Trust Hub on May 26, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/convert_to_markdown.py executes the pdftotext system utility and the pip module via subprocess.run. These calls use argument lists rather than shell strings, preventing command injection, and are restricted to specific, hardcoded operations for text extraction and dependency management.
  • [EXTERNAL_DOWNLOADS]: The skill supports the installation of optional Python extractors (pypdf, pdfplumber, PyMuPDF) from the official registry (PyPI) into a local virtual environment. This process is gated by the --install-missing flag and is used to provide the tool's core functionality.
  • [PROMPT_INJECTION]: The skill processes untrusted documents, creating an indirect prompt injection surface.
  • Ingestion points: PDF and DOCX files provided via the inputs argument in SKILL.md.
  • Boundary markers: None; document content is converted directly into Markdown without explicit isolation from instructions.
  • Capability inventory: Subprocess execution of pdftotext and file system write operations in scripts/convert_to_markdown.py.
  • Sanitization: The script employs a slugify function to sanitize output filenames, preventing directory traversal, and restricts processing to validated file extensions.
Audit Metadata
Risk Level
SAFE
Analyzed
May 26, 2026, 08:32 PM
Security Audit — agent-trust-hub — krt-document-forge