microsoft-forms-builder

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The PowerShell script scripts/browser_session.ps1 executes the Microsoft Edge binary (msedge.exe) using Start-Process. It configures the browser with specific flags (--remote-debugging-port and --user-data-dir) to enable programmatic control via the Chrome DevTools Protocol (CDP). This is the intended mechanism for the skill's automation.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes survey specifications from an external spec.json file, which constitutes an indirect prompt injection surface. The risk is assessed as safe because the data is treated as plain text for form fields.
  • Ingestion points: The build_form.py script reads and parses a user-provided JSON file containing the survey's title, description, and question text.
  • Boundary markers: The skill relies on standard JSON parsing; however, there are no specific markers or instructions to the agent to ignore potentially malicious instructions embedded within the survey text.
  • Capability inventory: The skill possesses capabilities for browser automation via Playwright (navigation, clicking, and typing), file system reading for the specification file, and file system deletion for cleaning up the temporary profile.
  • Sanitization: The script performs validation on the JSON schema to ensure data types and required fields are correct, but it does not sanitize the string content of survey questions or titles before they are entered into the Microsoft Forms UI.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 04:20 AM