Skill UI

Guides building reliable autonomous agents by covering agent loops, goal decomposition, reflection, and tooling patterns while stressing guardrails, human review, and auditability to curb compounding failure rates in production.

LlamaGuard is Meta’s 7–8B safety-specialized LLM that filters both prompts and responses by classifying six threat categories, enabling fast inference via vLLM/SageMaker and integration into NeMo Guardrails for end-to-end moderation.

NVIDIA’s NeMo Guardrails adds runtime safety rails to LLM apps, covering jailbreak detection, hallucination and fact-checking, PII filtering, and toxicity checks via Colang 2.0 rules and integrations for production use on T4 GPUs.

This guide implements comprehensive guardrails for Perplexity Sonar API. It covers crucial aspects like query content moderation, PII sanitization, enforcing user usage quotas, selecting appropriate models based on user tiers, and scoring citation reliability to ensure safe, ethical, and cost-effective AI usage.

This skill specializes in designing, optimizing, and evaluating prompts for Large Language Models (LLMs). It allows users to generate optimized prompt templates, structured output schemas (e.g., JSON mode, function calling), and comprehensive test suites. Use this skill when building new LLM applications, implementing complex reasoning patterns like Chain-of-Thought, defining system guardrails, or building robust evaluation frameworks to measure and improve model performance.

Meta’s Prompt Guard is an 86M-parameter classifier that spots prompt injections and jailbreak attempts before they reach LLMs, filtering user input, third-party data, or RAG documents with multilingual support and sub-2ms latency for fast deployments.

This skill provides implementation patterns for policy guardrails when utilizing Retell AI for voice agents and phone call automation. It is designed to help ensure that AI interactions in a telephony context adhere strictly to predefined business rules and safety policies, preventing unauthorized actions or deviations from the intended conversational flow.

This comprehensive script suite provides methods to optimize GPU cloud computing expenses on Vast.ai. It guides users through selecting the most cost-efficient GPU for their workload, leveraging interruptible (spot) instances, and implementing critical lifecycle management safeguards. Features include detecting idle compute resources, auto-destroy timers, and calculating total cost burn rates, ensuring AI model training and inference jobs stay within budget.

Evaluates incoming requests, decides if specialized skills are needed, orchestrates optimal combinations, and logs successful pairings in agent-memory while guarding against overusing complex tools on simple tasks.

An advanced reliability wrapper for AI agents that monitors, manages, and recovers from tool call failures. It automatically handles common issues such as truncated JSON responses, API timeouts, rate limits, and mid-chain process breakdowns. By implementing sophisticated failure classification, checkpointing, and learning mechanisms, Guardian ensures complex, multi-step agent workflows remain stable and resilient.

AuraKit converts a single /aura command into a secure full-stack Claude pipeline with 34 modes, OWASP SEC-01~15 safeguards, runtime hooks, tiered models, token savings, and multilingual reviewers.

This framework implements robust input and output validation guardrails for LLM-powered applications. It prevents prompt injection, detects and redacts PII, filters toxic content, and validates model outputs against defined schemas. Use it when deploying AI agents that handle sensitive data or require strict adherence to corporate policies, ensuring a secure and trustworthy AI interaction layer.