gbt-standard-docx

Pass

Audited by Gen Agent Trust Hub on Jun 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes external system commands soffice and pdftoppm to perform document conversion and rendering tasks.
  • These commands are invoked using list-based arguments within subprocess.run, which is a secure implementation that prevents shell injection.
  • [PROMPT_INJECTION]: The skill processes untrusted user data from standard document drafts, presenting a surface for indirect prompt injection.
  • Ingestion points: Data is read from user-supplied YAML/JSON files and existing Word documents through scripts/render_gbt.py and scripts/review_gbt_docx.py.
  • Boundary markers: User-provided paragraph text is interpolated into document templates and review results without explicit delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill possesses capabilities for filesystem writes and system command execution through its integrated scripts.
  • Sanitization: Format validation is applied to structured fields (e.g., standard numbers), but the textual content of paragraphs is not sanitized for potential instructional patterns.
  • [EXTERNAL_DOWNLOADS]: The skill documentation provides instructions for installation via npx skills, which involves downloading code and templates from the author's public repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 2, 2026, 09:24 AM
Security Audit — agent-trust-hub — gbt-standard-docx