The Agent Skills Directory

Dynamic Method Invocation: The fire command allows for calling named methods on live components via reflection. This provides significant flexibility for testing and automation but represents a surface for dynamic execution of application logic.
Runtime State Modification: The skill includes the ability to mutate dependency properties and XAML resources at runtime through the set-property and set-resource commands. This allows for direct alteration of the application's appearance and behavior during a session.
Indirect Prompt Injection Surface: The skill processes information from the application's visual tree, screenshots, and runtime logs. If the application under test displays or logs untrusted content, there is a possibility that the agent could encounter and follow instructions embedded within that data.
Ingestion points: The skill reads application state via tree, captures visual data via screenshot, and drains application output via logs.
Boundary markers: No specific delimiters or "ignore instructions" markers are defined for the data retrieved from the application.
Capability inventory: The skill possesses capabilities to interact with the UI (click, type, scroll), modify state (set-property), and execute methods (fire).
Sanitization: There is no evidence of sanitization or filtering applied to the application data before it is presented to the agent.
System Path Modification: The skill's deployment process, as described, involves modifying the user's PATH environment variable to include the tool's binaries. This is a common installation pattern for CLI utilities but constitutes a modification of the system's execution environment.

reactor-devtools