windows-dev-process-cleanup

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes built-in Windows command-line utilities to manage system processes.
  • Evidence: scripts/audit-dev-processes.ps1 uses cmd /c "taskkill /PID $RootPid /T /F" to terminate targeted process trees.
  • Evidence: scripts/audit-uwp-backgroundtasks.ps1 uses Stop-Process and tasklist /apps /fo csv /nh to identify and manage UWP applications.
  • Safety Controls: The scripts implement comprehensive verification logic, including comparing process creation timestamps and command-line fingerprints immediately before execution to ensure the target PID has not been reused by a different process.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from the local system's process list which could theoretically contain malicious strings.
  • Ingestion points: scripts/audit-dev-processes.ps1 reads process names and command lines using CIM (Win32_Process).
  • Boundary markers: The skill uses regex-based classification (e.g., matching npm outdated) and strict identity hashing rather than interpolating raw process strings into AI instructions.
  • Capability inventory: The primary capability is process termination (taskkill).
  • Sanitization: Validation is performed through deterministic PowerShell logic and identity checks rather than LLM-based filtering.
  • [PRIVILEGE_ESCALATION]: While the skill performs sensitive operations like process termination, it does not attempt to escalate its own privileges.
  • The scripts rely on the existing permissions of the user running the PowerShell environment and do not include commands like sudo, runas, or registry modifications to bypass User Account Control (UAC).
  • The documentation in references/migration-2.0.md explicitly notes that registry-based persistence methods for Phone Link were removed in version 2.0.0 in favor of standard Windows Settings.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 10:08 AM