wiztree

Warn

Audited by Gen Agent Trust Hub on Oct 24, 2026

Risk Level: MEDIUMPRIVILEGE_ESCALATIONPERSISTENCEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PRIVILEGE_ESCALATION]: The skill requires and instructs the agent to ensure high-privilege execution environment.
  • Evidence includes mandatory use of the /admin=1 flag for WizTree to perform raw NTFS MFT parsing.
  • Automation scripts in gpt.md are designed to run under the NT AUTHORITY\SYSTEM account with RunLevel Highest to bypass access restrictions.
  • The skill also provides commands for system-level modifications such as disabling hibernation (powercfg -h off) and purging the Windows component store (DISM).
  • [PERSISTENCE]: The skill documents and provides scripts for establishing persistence on the host system.
  • gpt.md contains a PowerShell script (Invoke-WizTreeAudit.ps1) and instructions to register it as a daily persistent task using Register-ScheduledTask in the Windows Task Scheduler.
  • [DATA_EXFILTRATION]: The skill includes functionality for sending disk analytics data to external services.
  • gpt.md provides code to dispatch storage alerts via webhooks using Invoke-RestMethod. While the provided URL is a placeholder, this mechanism is designed to exfiltrate drive capacity information and lists of top files to external endpoints (e.g., Discord/Slack).
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection through the processing of externally generated data.
  • Ingestion points: The skill reads CSV reports generated by WizTree using pd.read_csv in SKILL.md and Import-Csv in gpt.md.
  • Boundary markers: None identified; there are no explicit instructions to ignore embedded content in the ingested data.
  • Capability inventory: The skill has significant capabilities including executing shell commands (subprocess.run, Start-Process), managing scheduled tasks, and performing network requests (Invoke-RestMethod).
  • Sanitization: No evidence of data sanitization or validation of the CSV content before it is processed or used in subsequent logic.
  • [COMMAND_EXECUTION]: The skill generates and executes several system commands and scripts.
  • It utilizes subprocess.run in Python and Start-Process in PowerShell to interact with the underlying operating system and WizTree binaries.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 24, 2026, 04:40 AM