The Agent Skills Directory

[SAFE]: The skill serves as a guide for implementing the nemoguardrails library, which is an official NVIDIA framework designed to enhance LLM safety and reliability.
[PROMPT_INJECTION]: The skill contains strings typically associated with prompt injection attacks (e.g., "Ignore previous instructions", "Pretend you are DAN"). However, these are false positives as they are exclusively used within configuration examples to demonstrate how the guardrail system identifies and blocks malicious inputs.
[EXTERNAL_DOWNLOADS]: The skill identifies the nemoguardrails package and the meta-llama/LlamaGuard-7b model as dependencies. These are well-known resources provided by established technology organizations and follow standard installation practices.
[COMMAND_EXECUTION]: The skill includes a standard pip install command for the library. This is a routine operation for installing the documented tool and does not involve suspicious or obfuscated commands.

nemo-guardrails