skills/amitse/uiacli/win-uia/Gen Agent Trust Hub

win-uia

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill enables the agent to launch arbitrary executables and simulate system-wide mouse and keyboard inputs using low-level Win32 APIs.
  • Evidence: src/Uia.Core/ProcessManager.cs uses Process.Start to launch applications from user-provided paths. src/Uia.Core/InputSimulator.cs utilizes SendInput and SetCursorPos to simulate clicks, typing, and key combinations across the desktop.
  • [DATA_EXFILTRATION]: The skill provides tools to read potentially sensitive data from the system clipboard and capture full or partial desktop screenshots.
  • Evidence: src/Uia.Server/Program.cs exposes endpoints /clipboard (which calls System.Windows.Forms.Clipboard.GetText) and /screenshot (which uses Graphics.CopyFromScreen).
  • [EXTERNAL_DOWNLOADS]: The skill includes an installation one-liner that downloads a PowerShell script and binary assets from the author's GitHub repository.
  • Evidence: SKILL.md and README.md provide an installation command (irm ... | iex) targeting https://raw.githubusercontent.com/amitse/uiacli/master/install.ps1. The script fetches a ZIP archive containing the CLI and server binaries from the amitse/uiacli GitHub releases.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes UI automation data from external applications, which serves as an untrusted ingestion surface that could contain instructions for the agent.
  • Evidence:
  • Ingestion points: UI element names, values, and tree structures are retrieved in src/Uia.Core/ElementInspector.cs and returned to the agent.
  • Boundary markers: The skill does not explicitly utilize delimiters or 'ignore' instructions for data retrieved from application UI trees.
  • Capability inventory: The skill possesses the ability to launch processes, simulate input, and access the clipboard (src/Uia.Core/InputSimulator.cs, src/Uia.Core/ProcessManager.cs).
  • Sanitization: UI content is returned to the agent context as raw strings without validation or filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 02:45 AM
Security Audit — agent-trust-hub — win-uia