data-analyze-refined

Pass

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its core functionality of processing external data files.\n
  • Ingestion points: Data is ingested via pandas.read_csv and pandas.read_excel in src/data_analyzer.py.\n
  • Boundary markers: Absent. The skill does not implement delimiters or instructions for the AI agent to ignore directives embedded within the data rows.\n
  • Capability inventory: The skill has the capability to write local files (CSV, Excel, Word, HTML), execute Python sub-processes, and render content in a browser environment.\n
  • Sanitization: There is no explicit sanitization of data values before they are interpolated into the Jinja2 templates or Word documents generated in src/infographic_generator.py and src/main.py.\n- [COMMAND_EXECUTION]: The skill uses a Node.js entry point to orchestrate its Python-based analysis modules.\n
  • Evidence: src/index.js utilizes the python-shell package to execute main.py with user-supplied arguments.\n- [EXTERNAL_DOWNLOADS]: The skill's setup process involves downloading browser binaries from a well-known service to enable report capturing.\n
  • Evidence: The install.sh script executes npx playwright install chromium to download the browser required by the ScreenshotGenerator in src/screenshot_generator.py.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 12, 2026, 05:54 AM