The Agent Skills Directory

[SAFE]: The skill consists of technical documentation, architectural diagrams, and standard Dart/Flutter code templates. It promotes established best practices such as dependency injection, unidirectional data flow, and model transformation at layer boundaries.
[INDIRECT_PROMPT_INJECTION]: The skill defines templates for data layer clients (e.g., UserApiClient and LocalStorageClient) which process data from external REST APIs and local storage. This represents a standard application attack surface but does not introduce malicious intent to the agent itself.
Ingestion points: Code samples in packages/user_api_client/lib/src/user_api_client.dart and packages/local_storage_client/lib/src/local_storage_client.dart demonstrate fetching and decoding external data.
Boundary markers: The provided templates focus on architectural structure and use standard error handling; they do not include specific delimiters or prompt-injection warnings for the generated code, which is expected for general-purpose boilerplate.
Capability inventory: The templates use the http and shared_preferences packages for external data access.
Sanitization: The provided examples use standard JSON decoding and Dart type casting to handle external data structures safely.

layered-architecture