The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill architecture facilitates the processing of external data and user-provided instructions through its Agent Studio components.
Ingestion points: Data is ingested via the AgentWindow component for chat interactions and the DataService.ts for database synchronization.
Boundary markers: Explicit boundary delimiters for untrusted data are not specified in the provided documentation snippets.
Capability inventory: The skill includes a toolExecutor.ts for executing local frontend actions and facilitates communication with external LLM services via Server-Sent Events (SSE).
Sanitization: The skill implements a whitelist-based registration system (A2UI_REGISTRY) for dynamic UI rendering and uses structured OpenAI-format schemas for tool definitions to limit the model's action space.
[EXTERNAL_DOWNLOADS]: The project configuration references standard industry packages and platforms. All external dependencies and API integrations target well-known services like Supabase and Aliyun, which are consistent with the skill's stated purpose.
[CREDENTIALS_UNSAFE]: The documentation provides explicit instructions for secure secret management, advising developers to use the Supabase Secrets dashboard for the ALIYUN_BAILIAN_API_KEY rather than hardcoding it in source files or environment templates.

auto-develop