Skill UI

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.

Combines ESM3 and ESM C protein language models to generate sequences, predict structures, condition functions, and extract embeddings via local or Forge API inference for protein design and engineering workflows.

A comprehensive skill for fine-tuning and training large language and vision models using advanced techniques like TRL (SFT, DPO, GRPO) or Unsloth. This process runs entirely on managed Hugging Face cloud GPUs, eliminating the need for local GPU setup. It supports modern training workflows, model conversion (GGUF for local deployment), and ensures seamless artifact storage to the Hugging Face Hub.

An expert guide for deploying and optimizing Large Language Models (LLMs) entirely on local hardware. Mastering the entire local AI ecosystem, including inference engines like Ollama, vLLM, and llama.cpp. Provides deep knowledge on quantization formats (GGUF, EXL2, AWQ), VRAM calculation, and ensuring privacy-first, offline deployment for state-of-the-art models.

This skill generates a comprehensive issue report on the current challenges, limitations, and open problems within Japanese Natural Language Processing (NLP) for a specified topic. By combining knowledge from a curated local dataset with real-time web search, it provides a deep dive into ongoing research gaps, evaluating the state-of-the-art and recommending areas for future work.

A comprehensive studio workflow for generating professional, multilingual dubbing for local audio and video assets. This skill manages the entire process, including media submission, complex localization parameters (target locale, accent, language), real-time status polling, and final artifact download. It handles critical considerations like copyright checks, speaker consent, and preview generation to ensure high-quality, contextually appropriate dubbing across diverse languages.

A comprehensive studio for professional voice conversion. This tool allows users to transform local audio recordings into specific, authorized target voices. It manages complex workflows, requiring rigorous checks on source rights, target voice licensing, and detailed parameters like emotion, accent, pace, and language locale, ensuring compliance and high-quality output.

Use Voiceover Studio to convert written scripts into high-quality, professional audio narration. This tool supports diverse applications including product demos, IVR prompts, podcast episodes, and short-video voiceovers. It manages complex requirements such as specific accents, emotions, speaking pace, and precise language/locale preservation, ensuring the resulting audio is ready for immediate use.

A comprehensive guide and skill for deploying large language models (LLMs) locally using the llama.cpp framework and GGUF quantization format. It covers searching the Hugging Face Hub, selecting optimal quantizations (e.g., Q4_K_M), running local servers, and achieving OpenAI-compatible API endpoints. Supports various hardware backends including CPU, Mac Metal, CUDA, and ROCm. Ideal for developers needing offline, private LLM inference.

This skill evaluates and assembles the smallest, most efficient set of AI agent skills required to achieve a complex, natural language goal. It dynamically derives multi-step workflows, searching local indices, registries, and external sources using four specialized lenses (need, operation, outcome, method). It prioritizes user-friendly plain-language outputs while rigorously verifying compatibility, safety, and overlap between candidate skills before recommending the final stack.

This skill de-identifies sensitive PII from clinical notes across numerous languages, including Spanish, German, French, and Arabic. It automatically selects language-specific PII models, national ID regex patterns, and generates locale-aware fake surrogates. Designed for global healthcare datasets, it runs entirely on-device, ensuring maximum privacy and linguistic accuracy for diverse international clinical records.

Guides users to select a local, on-device Personally Identifiable Information (PII) detection model from a committed registry. It filters models based on language, required runtime environment (e.g., PyTorch, MLX), and size constraints. The process emphasizes building an offline shortlist and mandates rigorous pre-deployment recall validation and performance benchmarking to ensure secure, privacy-preserving deployment.