OpenSkillIndex← back to search
claude-codesecurity

Babysitter nemo-guardrails

NVIDIA NeMo Guardrails configuration for conversational safety and control

install
source · Clone the upstream repo
git clone https://github.com/a5c-ai/babysitter
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/a5c-ai/babysitter "$T" && mkdir -p ~/.claude/skills && cp -r "$T/library/specializations/ai-agents-conversational/skills/nemo-guardrails" ~/.claude/skills/a5c-ai-babysitter-nemo-guardrails && rm -rf "$T"
manifest: library/specializations/ai-agents-conversational/skills/nemo-guardrails/SKILL.md
tags
#guardrails#conversation-control#input-filtering#output-filtering#rail-configuration
source content

NeMo Guardrails Skill

Capabilities

  • Configure NeMo Guardrails rails
  • Design Colang conversation flows
  • Implement input/output rails
  • Set up topic control
  • Configure jailbreak detection
  • Implement fact-checking rails

Target Processes

  • system-prompt-guardrails
  • content-moderation-safety

Implementation Details

Rail Types

  1. Input Rails: Filter user inputs
  2. Output Rails: Filter LLM outputs
  3. Dialog Rails: Control conversation flow
  4. Retrieval Rails: Filter retrieved content
  5. Execution Rails: Control action execution

Colang Components

  • Flow definitions
  • Bot message templates
  • User message patterns
  • Actions and subflows

Configuration Options

  • Rails configuration
  • LLM selection
  • Embedding model
  • Action handlers
  • Custom rail implementations

Best Practices

  • Start with built-in rails
  • Design clear flows
  • Test with adversarial inputs
  • Monitor rail activations

Dependencies

  • nemoguardrails