The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes metadata flags derived from AI-generated responses to categorize conversations and monitor performance.
Ingestion points: The deriveCategory function in SKILL.md accepts AIResponseMetadata as input.
Boundary markers: The skill includes explicit PROHIBITED interface definitions and comments instructing developers to never store message content, user queries, or specific health-related topics.
Capability inventory: The skill utilizes database insert, update, and select operations for conversation tracking.
Sanitization: Content-based categorization is avoided by using high-level boolean flags (e.g., usedCrisisProtocol) to derive conversation categories.
[EXTERNAL_DOWNLOADS]: The skill provides links to external industry resources for metrics and best practices.
Sources: Documentation and blog posts from HiverHQ, Botpress, and Tidio.
Context: These links are used to provide background information on chatbot analytics best practices and key performance indicators.

chatbot-analytics