The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes a product architecture designed to ingest user-supplied voice data, which creates a potential surface for indirect prompt injection.
Ingestion points: references/detailed-guide.md describes the ingestion of user_speech via the AMAZON.SearchQuery slot.
Boundary markers: The documentation does not specify the use of delimiters or 'ignore' instructions for the processed user input.
Capability inventory: The architecture includes AWS Lambda execution, DynamoDB read/write access for conversation memory, and integration with the Anthropic Claude API.
Sanitization: No specific sanitization or filtering protocols for user-provided speech content are described in the architectural overview.
[EXTERNAL_DOWNLOADS]: The skill references several external resources and official SDKs for project implementation.
Evidence: Mentions and provides links for the Alexa Skills Kit (ASK) SDK for Python on GitHub (github.com/alexa/alexa-skills-kit-sdk-for-python) and Anthropic's Claude API documentation (docs.anthropic.com). These are recognized as well-known and trusted services.
[SAFE]: The skill documents best practices for security in its implementation plan, such as using AWS Secrets Manager for API keys and following LGPD (Brazilian data protection law) guidelines for user data retention.

auri-core