hexagone-web-feature-extractor

Pass

Audited by Gen Agent Trust Hub on Mar 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Instructs the agent to execute shell commands for installing the Playwright library and the Chromium browser binary.
  • [EXTERNAL_DOWNLOADS]: Downloads standard browser automation dependencies from the NPM registry and official browser repositories.
  • [DATA_EXFILTRATION]: Accesses an internal corporate server on a private network domain (ws004202.dedalus.lan) to extract application metadata and take screenshots of the ERP interface. These artifacts are stored on the local file system for report generation.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted text data (menu labels and page content) from the target web application. This data is interpolated into a metadata file and subsequently used to generate a Markdown document.
  • Ingestion points: Page titles, menu labels, and tab names extracted via page.evaluate() in SKILL.md.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the scraping logic.
  • Capability inventory: Uses page.screenshot() to capture images and fs.writeFileSync() via generate-md.js to write files to the local disk.
  • Sanitization: The generation script validates data types but does not sanitize content for Markdown or instruction injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 27, 2026, 09:25 AM