maintaining-windows-health

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Runtime workflow: the render-cleanup-plan.py HTTP server accepts outsider POST bodies to /submit (free-text JSON payload) and then materializes that into the selection JSON consumed by assets/apply-cleanup-selection.py.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill explicitly instructs installing software and services, registering scheduled tasks, and running elevated/irreversible cleanup commands (DISM, vssadmin, pnputil, PowerShell delete commands executed by the apply script), which modify system state and require elevation, so it pushes the agent to perform potentially system-compromising actions.