📦 mohitagw15856/pm-claude-skills

Supercharge Your Workflow with 174 AI Agent Skills

Are you ready to unlock the full potential of your favorite AI tools? This repository offers a massive collection of 174 professional agent skills designed to streamline everything from product management and engineering to design and business strategy. Instead of starting from scratch, you get access to pre-built slash commands and subagents compatible with platforms like Claude, ChatGPT, and Gemini. Whether you need to generate technical specs, conduct churn analysis, or build roadmaps, these ready-to-use workflows save you hours of effort. Dive into this professional toolkit today and transform how you work with AI to become significantly more productive.

📰 https://news.ycombinator.com/item?id=48600385

🆔 @hackernewsgithubprojects

📦 jaytel0/taste

Turn Reference Images Into AI Design Skills

Ever wish your AI could consistently capture a specific visual style? This repository provides a powerful pipeline that transforms your own reference images into a reusable design skill. Instead of relying on vague aesthetic prompts, this system analyzes your images to generate concrete, actionable rules for layout, spacing, and typography. By fusing insights from multiple vision models, it creates a structured guide that helps AI agents produce more accurate, professional results. Whether you are building interfaces or dashboards, you can now enforce specific design constraints automatically. Simply feed your images into the pipeline to start building your own custom design library.

🆔 @hackernewsgithubprojects

📦 thu-si/spatial-ttt

Mastering Spatial Intelligence with Spatial-TTT

Have you ever wondered how AI can better understand the world through streaming video? Meet Spatial-TTT, an innovative framework designed to tackle long-horizon visual spatial tasks. By combining a hybrid architecture with test-time training, this method creates compact memory that accumulates crucial 3D spatial evidence from continuous video streams. It uses a smart spatial-predictive mechanism and sliding-window attention to effectively organize visual signals over time. This repository provides the official implementation, training scripts, and evaluation tools to help you explore state-of-the-art performance in spatial intelligence. Dive in today to see how this approach is redefining video understanding.

🆔 @hackernewsgithubprojects

📦 igoraherne/trellis.2-stableprojectorz

Generate High-Fidelity 3D Assets with TRELLIS.2

Ever wanted to turn a single image into a detailed 3D model in seconds? This repository integrates the powerful TRELLIS.2 generative model, allowing you to create high-fidelity 3D assets with complex textures and sharp geometry effortlessly. By utilizing a unique sparse voxel structure called O-Voxel, it handles tricky shapes like open surfaces and non-manifold geometry far better than traditional methods. The project is optimized to run on consumer GPUs, turning simple images into PBR-ready 3D files for your projects. Dive into this cutting-edge 3D generation toolkit today to start bringing your visual ideas to life with professional-grade efficiency.

🆔 @hackernewsgithubprojects

📦 mukundbala/rolling-bonxai

Efficient Large-Scale Mapping in ROS 2 with Rolling-Bonxai

Are you struggling with memory limits while mapping massive areas with your robots? Rolling-Bonxai is a clever solution that optimizes voxel grid usage for ROS 2. Instead of keeping everything in memory, it implements a chunking system that loads and offloads voxel maps based on the robot's movement. It efficiently saves unused map parts to your disk and retrieves them only when needed, significantly reducing the memory footprint. With built-in policies for chunk management and asynchronous I/O, it keeps your navigation smooth even in large-scale environments. Streamline your robot's mapping performance today and keep your system memory-efficient while exploring.

🆔 @hackernewsgithubprojects

📦 ulsklyc/yuvomi

Take Control of Your Household with Yuvomi

Are you tired of relying on big tech to manage your family life? Meet Yuvomi, a powerful self-hosted family planner that puts your data directly into your own hands. This private, offline-capable platform lets you organize tasks, calendars, shopping lists, meal plans, and even household budgets all in one beautiful interface. It runs effortlessly on your home server or NAS using Docker, offering a polished, mobile-first experience without any subscription fees or hidden trackers. Whether you need to track shared expenses or plan weekly meals, Yuvomi gives you full control. Reclaim your digital privacy and simplify your home management today.

🆔 @hackernewsgithubprojects

📦 raidowl/homelab-hub

Take Control of Your Home Lab Infrastructure

Are you tired of losing track of your servers, network hardware, and services? Home Lab Hub is a powerful self-hosted web application designed to help you inventory and visualize your entire home lab setup in one centralized location. It provides full management for your virtual machines, storage devices, and network shares while offering an interactive graph map to see how everything connects. You can even document your projects using the built-in markdown editor. Whether you are running a complex cluster or just starting out, this tool makes infrastructure tracking effortless. Give your lab the organization it deserves today.

🆔 @hackernewsgithubprojects

📦 ngrok/webernetes

Run Kubernetes Entirely in Your Browser

Have you ever wanted to experiment with Kubernetes without the headache of managing complex cluster infrastructure? Meet this project, which brings a subset of Kubernetes directly into your browser. By acting as a simulator, it removes the need for any backend server components, allowing you to boot up clusters instantly on your local machine. It uses custom implementations of controllers and internal resource handling to replicate cluster logic without needing real container images. It is an incredibly clever way to build interactive tutorials and educational content, making Kubernetes experimentation accessible, lightweight, and truly portable for everyone.

📰 https://news.ycombinator.com/item?id=48588578

🆔 @hackernewsgithubprojects

📦 miracodeai/mira

Stop Paying For AI Code Reviews: Meet Mira

Tired of expensive AI code reviewers that lock your data into their cloud? Meet Mira, a self-hosted AI code review platform that lets you keep your code, configurations, and dashboards entirely on your own infrastructure. Unlike standard SaaS tools, Mira builds a comprehensive index of your entire codebase, allowing the AI to reason about call sites, dependencies, and project patterns rather than just analyzing the diff. It includes powerful features like automated vulnerability scanning, org-wide package tracking, and a smart learning loop that adapts to your team's specific standards. Take control of your developer workflow and secure your code today.

📰 https://news.ycombinator.com/item?id=48570197

🆔 @hackernewsgithubprojects

📦 kennss/siliconscope

Stop Guessing: Monitor Your Apple Silicon Mac Properly

Ever wonder what is actually happening inside your Mac when you run local AI models or heavy media tasks? SiliconScope is a native system monitor built specifically for Apple Silicon that gives you the deep visibility Activity Monitor hides. It tracks your Neural Engine, media engines, and memory bandwidth in real-time without needing any sudo permissions. You can spot thermal throttling, identify performance bottlenecks, and even measure your model speed in tokens per second. It is the perfect tool for power users who want to understand their hardware efficiency. Grab the latest version and start monitoring your Mac like a pro today.

🆔 @hackernewsgithubprojects

📦 turbo-tan/llama.cpp-tq3

Run Massive AI Models on Consumer GPUs!

Ever wanted to run powerful, large-scale AI models on your own computer without needing an enterprise-grade setup? This project is a specialized version of the popular inference engine designed to make that dream a reality. By using advanced Walsh-Hadamard transform quantization, it shrinks model sizes by ten percent while maintaining high quality. This breakthrough allows you to fit 27-billion parameter models onto standard 16-gigabyte graphics cards with impressive speed. Whether you are an AI enthusiast or a developer, this tool is your key to unlocking greater performance and efficiency on your local hardware. Start experimenting with more capable models today.

🆔 @hackernewsgithubprojects

📦 kaopu-xiaopu/web-design

Design Better Websites Faster with This AI Skill

Stop jumping straight into messy code and start designing with intent. This repository provides a specialized skill for Claude Code that flips your workflow by putting the design specification first. By feeding it a simple prompt, a URL, or a screenshot, you generate a structured design document covering everything from color palettes to accessibility rules before a single line of code is written. It acts as a professional blueprint, ensuring your web projects remain consistent and portable. Whether you are building complex interfaces or sleek landing pages, this tool helps you maintain high quality throughout the development process.

🆔 @hackernewsgithubprojects

📦 goekdeniz-guelmez/mlx-lora-studio

Train LLMs Locally on Your Mac with MLX LoRA Studio

Fine-tuning large language models usually feels like a chore, but what if you could do it right on your Mac with just one click? MLX LoRA Studio is a native macOS application designed to make training fully on-device, private, and accessible. It handles all the complex environment setups and dependencies for you, providing a clean interface to run algorithms like SFT, DPO, and QLoRA. You can monitor live metrics, generate synthetic data, and even upload your finished adapters to Hugging Face without leaving the app. Ready to start training models on your own machine, keep your data private, and simplify your workflow?

🆔 @hackernewsgithubprojects

📦 mozilla-ai/cq

Stop Your AI From Repeating Mistakes With This Shared Learning Tool

Tired of your AI coding agents making the same mistakes over and over? Meet this project, an open standard designed to help agents share collective knowledge so they stop rediscovering the same failures independently. It works by allowing your agents to query a store for proven solutions before they act, or propose new knowledge units when they uncover undocumented behaviors or workarounds. By enabling persistent, shared memory across sessions and machines, this tool helps your agents learn from each other in real time. It is a smart way to build a smarter, more efficient development workflow. Check it out to start evolving your agents today.

🆔 @hackernewsgithubprojects

📦 ayushh0110/screenmind

ScreenMind: Your Local Private AI Memory

Ever wish you had a photographic memory for your digital life? Meet ScreenMind, an open-source project that captures your screen and turns it into a searchable AI database. Unlike other tools that compromise your data, this runs entirely offline using Gemma 4, ensuring your activity stays private on your own machine. It intelligently tracks screen changes, records meetings, and lets you chat with your past activity to find exactly what you need. With features like sensitive data redaction and encrypted storage, it is the ultimate tool for reclaiming your digital history. Take control of your workflow with local AI intelligence.

📰 https://news.ycombinator.com/item?id=48522417

🆔 @hackernewsgithubprojects

📦 jangafx/fmag

GPU Programming with Only One Instruction

Have you ever wanted to run complex programs on a GPU without the nightmare of branch divergence? Meet FMAG, a fascinating virtual machine and toolchain built on a single-instruction architecture. By replacing traditional complex branching with a clever fused multiply-add and guard operation, it executes programs in lockstep across all GPU lanes, completely avoiding the performance traps of divergent control flow. This C library lets you build, optimize, and assemble custom instruction streams that are highly efficient on modern hardware. It is a brilliant solution for running runtime-generated content on the GPU, proving that sometimes the simplest ideas are the most powerful.

📰 https://news.ycombinator.com/item?id=48573342

🆔 @hackernewsgithubprojects

📦 stevethekiller/killerpdf

The Best Portable PDF Editor for Windows

Are you tired of bloated PDF software that demands subscriptions and tracks your every move? Meet KillerPDF, a lightweight, local-only editor designed specifically for Windows. It provides all the essential tools field technicians and power users need without the annoying account requirements or background telemetry. You can easily view, annotate, merge, split, and even sign documents or fill out interactive forms directly on your machine. Whether you need a portable tool to run from a drive or a permanent installation, it offers high-quality rendering and total control over your files. Take back your workflow with this simple, efficient PDF solution.

🆔 @hackernewsgithubprojects

📦 leoyeai/openclaw-marketing-skills

Boost Your Marketing Results with 33 Proven Agent Skills

Unlock a powerhouse of growth for your business with this collection of thirty-three battle-tested marketing skills designed for OpenClaw agents. This repository acts as a comprehensive toolkit, providing structured guidance on everything from conversion rate optimization and SEO audits to copywriting frameworks and paid ad strategies. Instead of guessing, you can use these resources to implement data-driven tactics like email sequences, churn prevention, and product marketing context setup. It is the perfect way to automate and sharpen your marketing execution by leveraging ready-to-use professional playbooks. Start mastering your marketing strategy and see real results in your workflow today.

🆔 @hackernewsgithubprojects

📦 hwdsl2/self-hosted-ai-stack

Build Your Own Private AI Stack with Docker

Ever wanted to run your own powerful AI tools without relying on the cloud? This project lets you deploy a complete, self-hosted AI stack using just a single command with Docker Compose. You get full control over local LLMs through Ollama, a user-friendly chat interface with AnythingLLM, and powerful features like real-time voice transcription and document parsing all in one private environment. It is designed to be secure, supports GPU acceleration for faster performance, and is easy to scale up or down. Take charge of your data and start building your own local AI infrastructure today.

📰 https://news.ycombinator.com/item?id=48121643

🆔 @hackernewsgithubprojects

📦 pseudoku/kotinos

Custom Ergo Fingertip Shell Mod for HSK Pro

Ever wanted to transform your gaming mouse into a truly custom ergonomic masterpiece? Meet Kotinos, a unique open-source project that provides 3D printable files to create a custom fingertip shell mod for the G-Wolves HSK Pro mouse. This project includes STL files for the shell and paddles, plus detailed assembly instructions for using your own components like ribbon cables and connectors. You can even use the provided SCAD files to tweak the geometry to perfectly match your hand measurements. It is a fantastic way to take full control of your hardware comfort. Grab the files and start building today.

📰 https://news.ycombinator.com/item?id=48491385

🆔 @hackernewsgithubprojects