The Agent Skills Directory

[SAFE]: The skill is an import from the NVIDIA RAG Blueprint, adhering to standard DevOps and AI evaluation practices without malicious intent.
[COMMAND_EXECUTION]: The skill guides the user to run evaluate_rag.py using the uv toolchain to perform ingestion and evaluation against a RAG server.
[EXTERNAL_DOWNLOADS]: Instructions include the materialization of documents from public links into a local corpus/ directory to facilitate retrieval testing.
[CREDENTIALS_UNSAFE]: The skill utilizes the NVIDIA_API_KEY for RAGAS scoring but provides comprehensive security guidance on avoiding shell history exposure and preventing the commitment of secrets to version control.

rag-eval