Classify LLM inputs and outputs into six safety categories.
npx skills add https://github.com/ricable/mcai --skill llamaguard-ricable
Please help me install this Agent Skill. Skill: llamaguard Source: https://github.com/ricable/mcai/tree/main/.agents/skills/llamaguard Command: npx skills add https://github.com/ricable/mcai --skill llamaguard-ricable
LlamaGuard helps you catch unsafe user prompts and risky model outputs before they reach production, reducing exposure to harmful, disallowed, or policy-violating content.
Use the llamaguard skill to evaluate a chat message and decide whether it should be allowed, blocked, or escalated.
High-intent search queries and answers about installing and using this skill.
Content moderation guardrails block unsafe LLM inputs and outputs by classifying chat messages across six safety categories, allowing you to allow, block, or escalate risky conversations before they reach production.
LLM safety filtering is the process of screening chatbot conversations to detect policy-violating content. It classifies messages into safe or unsafe outcomes to prevent high-risk topics like self-harm guidance from reaching users.
Yes, this content moderation approach works with common deployment paths including vLLM, Sagemaker, Transformers, FastAPI, and NeMo Guardrails, ensuring deterministic safety screening across different API-backed LLM applications.
You apply input and output filtering by evaluating both user prompts and model responses against six safety categories, blocking requests for harmful content while allowing normal conversations to flow through your API-backed LLM assistant.
Yes, you can use NeMo Guardrails to enforce deterministic moderation behavior. This compatibility ensures your LLM application applies consistent safety screening rules across all processed chat messages.
Six-category safety detection focuses strictly on predefined unsafe outcomes like weapons or self-harm. It does not catch content outside these specific categories, meaning normal conversations flow through but nuanced policy violations might be missed.
ovachieverClassify user input and system outputs across six safety categories.
shogo-labsEscalate critical incidents to designated team channels with structured alert messages.
skywatch-bskyInvestigate suspicious AT Protocol accounts using a six-phase methodology.
XerrionIdentify and document LLM prompt-injection risks and tool-call abuse in AI systems.
