hqq-quantization
davila7Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.
quantizing-models-bitsandbytes
davila7Quantizes LLMs to 8-bit or 4-bit for 50-75% memory reduction with minimal accuracy loss. Use when GPU memory is limited, need to fit larger models, or want faster inference. Supports INT8, NF4, FP4 formats, QLoRA training, and 8-bit optimizers. Works with HuggingFace Transformers.
llama-cpp
davila7Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
llama-factory
davila7Expert guidance for fine-tuning LLMs with LLaMA-Factory - WebUI no-code, 100+ models, 2/3/4/5/6/8-bit QLoRA, multimodal support
gptq
davila7Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.
awq-quantization
davila7Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss. Use when deploying large models (7B-70B) on limited GPU memory, when you need faster inference than GPTQ with better accuracy preservation, or for instruction-tuned and multimodal models. MLSys 2024 Best Paper Award winner.
cryptowallet
openclawComplete cryptocurrency wallet management for Web3, DeFi, and blockchain applications. Create and manage EVM (Ethereum, Polygon, BSC, Arbitrum, Optimism, Base, Avalanche) and Solana wallets with encrypted local storage. Query balances for native tokens (ETH, MATIC, BNB, SOL) and standard tokens (ERC20, SPL). Send transactions, interact with smart contracts, and manage multiple addresses across 12+ networks. Secure password-protected key storage with AES-256 encryption. Use for: (1) Creating new crypto wallets, (2) Importing existing wallets, (3) Checking token balances across chains, (4) Sending cryptocurrency and tokens, (5) Interacting with DeFi protocols and smart contracts, (6) Multi-chain portfolio management, (7) NFT transfers, (8) Blockchain development and testing. Keywords: crypto, cryptocurrency, wallet, blockchain, ethereum, solana, web3, defi, token, erc20, nft, smart contract, metamask alternative, hardware wallet, cold storage, hot wallet, blockchain wallet, digital wallet, bitcoin.
office-sprite
paulrobelloGenerate game-ready sprites for the Claude Office Visualizer using Nano Banana MCP and ImageMagick. Uses a 45-degree front/top-down perspective (NOT isometric) with a retro 16-bit pixel art style. Includes a validation step before processing and helper scripts for background removal.
wendy-contributing
JoannisExpert guidance on contributing to WendyOS: Yocto builds, agent internals, E2E testing, and system architecture. Use when developers mention: (1) building WendyOS images, (2) meta-wendyos layers or bitbake, (3) wendy-agent development or internals, (4) containerd or nerdctl on WendyOS, (5) E2E tests for wendy-agent, (6) Yocto recipes or bbappend files, (7) mDNS/Avahi service configuration, (8) device identity or UUID generation.
Frontend UI/UX Engineer
404kidwizA designer-turned-developer who crafts stunning UI/UX even without design mockups. The code can be a bit messy, but the visual output is always outstanding.
model-quantization
Expert capability for AI model quantization and optimization. Covers 4-bit and 8-bit quantization, GGUF conversion, memory optimization, and quality–performance tradeoffs for deploying large language models in resource-constrained JARVIS environments.
prompt-optimization-claude-45
Jamie-BitFlightOptimize CLAUDE.md files and Skills for Claude Code CLI. Use when reviewing, creating, or improving system prompts, CLAUDE.md configurations, or Skill files. Transforms negative instructions into positive patterns following Anthropic's official best practices.
Git Master
JosiahSiegelComplete Git expertise system for ALL git operations. PROACTIVELY activate for: (1) ANY Git task (basic/advanced/dangerous), (2) Repository management, (3) Branch strategies and workflows, (4) Conflict resolution, (5) History rewriting/recovery, (6) Platform-specific operations (GitHub/Azure DevOps/Bitbucket), (7) Advanced commands (rebase/cherry-pick/filter-repo). Provides: complete Git command reference, safety guardrails for destructive operations, platform best practices, workflow strategies, reflog recovery techniques, and expert guidance for even the most risky operations. Always asks user preference for automatic commits vs manual control.
PowerShell Master
JosiahSiegelComplete PowerShell expertise system across ALL platforms (Windows/Linux/macOS). PROACTIVELY activate for: (1) ANY PowerShell task (scripts/modules/cmdlets), (2) CI/CD automation (GitHub Actions/Azure DevOps/Bitbucket), (3) Cross-platform scripting, (4) Module discovery and management (PSGallery), (5) Azure/AWS/Microsoft 365 automation, (6) Script debugging and optimization, (7) Best practices and security. Provides: PowerShell 7+ features, popular module expertise (Az, Microsoft.Graph, PnP, AWS Tools), PSGallery integration, platform-specific guidance, CI/CD pipeline patterns, cmdlet syntax mastery, and production-ready scripting patterns. Ensures professional-grade, cross-platform PowerShell automation following industry standards.
qlora
itsmostafaMemory-efficient fine-tuning with 4-bit quantization and LoRA adapters. Use when fine-tuning large models (7B+) on consumer GPUs, when VRAM is limited, or when standard LoRA still exceeds memory. Builds on the lora skill.
mlx-apple-silicon
plurigridRun LLMs on Apple Silicon with MLX/mlx_lm - unified memory, 4-bit quantization, streaming generation, prompt caching. Optimal for M-series chips.
llama-cpp
AXGZ21Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.