requirements-analysis

Pass

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-supplied requirement documents, creating an attack surface for indirect prompt injection.
  • Ingestion points: Untrusted requirements are ingested into the AI context via the system prompt defined in prompts/requirements-analysis.md.
  • Boundary markers: The prompt employs Markdown headers and explicit role definitions (Lead QA Engineer) to separate instructional instructions from the user data.
  • Capability inventory: The skill includes scripts for reading and writing files and converting them into multiple formats (JSON, CSV, DOCX, XLSX, XMind) via Python.
  • Sanitization: The conversion utility scripts/convert_formats.py uses standard parsing libraries (json, csv, xml.etree.ElementTree) which are robust for structured data parsing but do not specifically filter for malicious natural language instructions embedded within the requirements text.
  • [COMMAND_EXECUTION]: The skill contains several wrapper scripts (e.g., scripts/convert_to_csv.py, scripts/batch_convert_templates.py) that utilize subprocess.call to invoke the core conversion logic in scripts/convert_formats.py. These executions are performed using the local Python interpreter and are restricted to internal skill files, representing legitimate tool functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 12, 2026, 06:17 PM