manual-testing-en

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-provided testing requirements and external files (Markdown, JSON, CSV, DOCX, XLSX, XMind) which serve as ingestion points for untrusted data.
  • Ingestion points: File parsing logic in scripts/convert_formats.py and scripts/parse_formats.py.
  • Boundary markers: The system prompt in prompts/manual-testing_EN.md accepts requirements and context directly without explicit delimiters or instructions to disregard embedded instructions within the user-provided data.
  • Capability inventory: The skill can execute local Python scripts via subprocess and perform file system read/write operations.
  • Sanitization: There is no evidence of input validation or secure XML parsing; the code uses the standard xml.etree.ElementTree library, which is susceptible to XML-based attacks if used on malicious files.
  • [COMMAND_EXECUTION]: The skill employs internal command execution to chain its processing scripts.
  • Evidence: scripts/batch_convert_templates.py, scripts/convert_output_formats.py, and several conversion wrapper scripts (e.g., scripts/convert_to_json.py) use subprocess.call to execute other Python scripts within the package using sys.executable.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 03:11 AM