human-priority-desktop-control
Pass
Audited by Gen Agent Trust Hub on Jun 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The Python helper script (
scripts/human_priority_desktop.py) executes a bundled PowerShell script (scripts/desktop_uia.ps1) to interface with the Windows UI Automation API. This is a core part of its documented functionality to discover and interact with native Windows controls. - [PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it reads and processes data from application UI elements (names, automation IDs, text content).
- Ingestion points: UI element property extraction occurs in
scripts/desktop_uia.ps1and is presented to the agent viauia-treeanduia-findcommands. - Boundary markers: The documentation defines a strict 'AI Pre-Review + Human Final Review' workflow, instructing the agent to never treat its own summary as final approval.
- Capability inventory: The skill allows full desktop manipulation, including input injection via
SendInput, screenshot capture, and clipboard access through Win32 APIs inscripts/human_priority_desktop.py. - Sanitization: No explicit sanitization of UI text is performed; risk is mitigated by the low-level keyboard/mouse hooks that pause automation if the human user takes over.
- [SAFE]: No malicious patterns, obfuscation, or unauthorized network operations were identified. The skill's implementation aligns with its stated purpose and incorporates extensive safety instructions to minimize automation risks.
Audit Metadata