The Agent Skills Directory

Indirect Prompt Injection Surface: The skill instructs the agent to fetch and research external documentation from sources such as the official MAUI repository and platform API docs. This creates a surface where external content could theoretically influence agent behavior.
Ingestion points: Untrusted data may enter the context via web_fetch calls for MAUI source code and platform API documentation as described in SKILL.md.
Boundary markers: There are no specific delimiters or instructions provided to the agent to isolate external content from its internal logic.
Capability inventory: The skill involves generating code capable of UI interaction (TryNativeTap), capturing screenshots (CaptureScreenshotAsync), and utilizing reflection as outlined in devflow-integration.md and extensibility-gaps.md.
Sanitization: The skill does not define specific sanitization or validation protocols for the external documentation it retrieves.
Dynamic Execution via Reflection: The documentation in extensibility-gaps.md and the anti-patterns section of SKILL.md describe the use of reflection and DispatchProxy to access internal MAUI APIs (such as SetDefault for Essentials services and AlertManager for dialogs). While these involve dynamic access to members, they are characterized as necessary architectural workarounds for existing framework limitations and are intended for implementation within the project structure.

maui-platform-backend