The Agent Skills Directory

[SAFE]: Analysis of the skill instructions and documentation reveals no malicious patterns, prompt injections, or unauthorized data access attempts. The skill operates as a read-only directory.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs users on how to install the exiryab-mcp package using the uvx tool. This is a standard deployment method for Model Context Protocol (MCP) servers and the package is specific to the service described.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge for user queries and external API data, creating a potential surface for indirect injection.
Ingestion points: User search terms (name, city, specialty) in SKILL.md and data returned by the ey_search_doctors tool.
Boundary markers: The instructions define confidence levels (high, medium, low) to guide the agent's interpretation of external data.
Capability inventory: Tools for searching doctors (ey_search_doctors), listing cities (ey_cities), and specialties (ey_specialties).
Sanitization: Instructions include logic for handling warnings and dropped filters to ensure the agent communicates limitations to the user.

exiryab