The Agent Skills Directory

Indirect Prompt Injection Surface: The skill allows the agent to ingest external data from the application under debug, such as application logs via maui devflow ui logs and WebView DOM snapshots via maui devflow webview snapshot. This data originates from the application and is not sanitized, creating a surface where malicious application content could potentially influence agent behavior. Boundary markers are currently absent in these ingestion points.
Command Execution: The skill facilitates the execution of various system commands for application building (dotnet build), deployment, and device management using tools like adb and xcrun simctl. These capabilities are necessary for development but grant the agent significant control over the local shell environment.
Privilege Escalation: The Linux/GTK platform documentation includes instructions for using sudo to install system dependencies (xdotool, ydotool) and to modify user group memberships and udev rules for hardware access. While intended for environment configuration, these actions involve elevated permissions.
Sensitive Data Access: The skill provides commands to read and modify application-specific state, including maui devflow preferences and maui devflow secure-storage. This grants the agent access to potentially sensitive information stored within the context of the application being debugged.

maui-ai-debugging