The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides templates and instructions that ingest untrusted data from external sources (Telegram API) and process it within the agent context.
Ingestion points: In references/quickstart.md, the handle_message function accepts types.Message objects, specifically accessing the message.text attribute.
Boundary markers: The quickstart example lacks delimiters or explicit instructions for the model to treat the message.text content as data rather than instructions.
Capability inventory: As documented across references/fsm.md and references/quickstart.md, the framework supports message responses, complex state management (FSM), and potential integration with external storage like Redis or MongoDB.
Sanitization: The skeleton code await message.answer(message.text or "") directly reflects user-supplied content without validation or sanitization, creating a surface where malicious inputs could influence the bot's behavior.

aiogram-framework