Skill UI

A comprehensive toolkit designed to manage the complex data flow associated with Perplexity search workflows. This skill implements critical security measures, including PII sanitization in user queries, validates external citations for accuracy, optimizes performance via result caching with freshness policies, and manages conversation context to maintain coherence and reliability. Essential for building robust, compliance-aware AI applications.

Covers prompt and response caching plus Cache Augmented Generation patterns for LLMs such as Claude and OpenAI to reuse stable context, reduce repeat work, and cut latency/cost on recurring queries.

Designs and implements production-grade RAG systems with document chunking, embeddings, vector stores, hybrid search, reranking, and evaluation, ideal for knowledge-grounded AI apps that need semantic retrieval, similarity search, and context augmentation.

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.

RWKV is an RNN+Transformer hybrid architecture with linear-time inference, infinite context, and GPT-style training. It delivers constant-memory, streaming generation on CUDA and CPU, ideal for long-context, memory-constrained deployments.

Provides a structured workflow for defining agent-agnostic subagents with isolated context, covering purpose definition, metadata configuration, prompt writing, and common specialist patterns for verification, debugging, security auditing, and review.

Provides production-ready Megatron-Core workflows for training 2B–462B parameter LLMs with tensor/pipeline/expert/context parallelism, maximizing GPU efficiency on H100 clusters.

Vexor is a powerful command-line interface designed for advanced semantic search across local files. By utilizing vector embeddings and integrating with large language models (like Claude/Codex), it allows users to search files based on meaning and context, rather than just exact keywords. This is ideal for large codebases or document repositories where traditional keyword search methods are insufficient.

Reasoning-Driven Image Briefs skill transforms agent prompts into logic-based Gemini 3 style instructions for muapi.ai, adding structured creative briefs (subject/action/context/composition/lighting), precise text cues, and physics-grounded constraints before invoking generation scripts.

Self-hosted open-source alternative to NotebookLM for organizing research, ingesting diverse sources, generating AI notes, running context-aware chats, performing vector search, producing multi-speaker podcasts, and automating workflows via a REST API.

Hierarchical Agent Memory (HAM) scopes Claude Code agent context by generating per-directory CLAUDE.md files, tracking token savings, and surfacing analytics via a dashboard so multi-folder projects avoid repeated full scans.

Helps product managers diagnose whether AI workflows rely on indiscriminate context stuffing or deliberate context engineering, then apply bounded domains, retrieval, and Research→Plan→Reset→Implement practices to improve reliability and output quality.