The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to interact with device simulators, manage processes, and process video files.
Utilizes xcrun simctl and adb shell screenrecord for capturing test journeys on iOS and Android as detailed in references/flutter.md.
Uses kill commands for process control and cleanup after recording operations.
Employs ffprobe and ffmpeg for media analysis and frame extraction as part of the verification step.
[EXTERNAL_DOWNLOADS]: The skill utilizes the Dart ecosystem for its testing infrastructure.
Executes the marionette_mcp package via dart run as described in the server registration section of references/flutter.md.
References the marionette_flutter package as a project dependency required for app binding.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data sources that could potentially harbor malicious instructions, representing an inherent vulnerability surface for E2E testing tools.
Ingestion points: The skill reads application logs, UI element metadata, and visual artifacts through tools like get_logs, get_interactive_elements, and ffmpeg frame extraction.
Boundary markers: The instructions do not specify explicit delimiters or protective instructions for the agent when interpreting external log or UI content.
Capability inventory: The agent has capabilities including shell command execution (dart, xcrun, adb, ffmpeg, kill) and local file system access.
Sanitization: No specific sanitization or filtering of log content or UI text is described before the data is processed by the agent.

e2e