The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from third-party API capabilities discovered and executed through the QVeris MCP server, creating an attack surface for indirect instructions.
Ingestion points: The agent ingests external content via the discover, inspect, and call MCP tools, as well as responses from the generated production code targeting qveris.ai.
Boundary markers: The instructions lack explicit boundary markers or system instructions to ignore potential command patterns embedded in tool outputs or API responses.
Capability inventory: The skill utilizes tool discovery and execution capabilities, and generates Python code capable of performing authenticated network operations.
Sanitization: There are no defined procedures for sanitizing or validating the structure of external content to prevent the agent from interpreting data as instructions.
[SAFE]: The skill references vendor-owned resources and well-known services.
Fetches guidelines from the official QVerisAI GitHub repository.
Communicates with the official qveris.ai API domain.
Uses the well-known requests Python library for network operations.

qveris