📦 forcedotcom/sf-skills

Salesforce Skills Library

Salesforce just released a massive, ready-made library of specialized instructions that teach AI agents how to build real applications for their platform. This project, called the Salesforce Skills Library, solves the problem of vague coding suggestions by providing strict, tested workflows for everything from backend logic to user interface components. Instead of guessing, your AI assistant follows these precise blueprints to write code that actually fits the ecosystem. It works with popular tools like Claude Code and Cursor, meaning you can plug it in right now. The library is actively evolving, so it reflects the current best practices directly from the experts.

🆔 @hackernewsgithubprojects

📦 ahmadrosid/nakama

Nakama: Your Team's AI Agent Platform

Nakama is the AI agent platform built specifically for teams to work alongside human employees. It gives every AI agent its own identity, permissions, and memory, so they can handle tasks across Telegram, WhatsApp, and Discord without breaking the fourth wall. The coolest part is that each agent has a soul, meaning it remembers context and maintains a consistent personality across different channels and sessions. It is not just a chatbot; it is a structured environment where multiple organizations can coexist on a single server with strict security boundaries. This setup turns loose AI scripts into reliable, manageable digital teammates that respect your team structure.

🆔 @hackernewsgithubprojects

📦 flashml-org/freetoken

Run Frontier AI on Your Gaming PC

FreeToken is the local inference engine that lets your gaming PC run massive AI models. It treats your computer’s memory and processor as one flexible system, so you can serve huge models without expensive server racks. The magic is in its smart memory swapping, which automatically moves parts of the AI brain between your fast graphics card and slower system memory to keep things running smoothly. This means you can chat with powerful, private AI assistants directly on your own hardware, bypassing the need for cloud subscriptions or massive electricity bills. It is a huge leap for developers who want full control over their tools.

📰 https://news.ycombinator.com/item?id=49394148

🆔 @hackernewsgithubprojects

📦 bagelhole/devops-security-agent-skills

The AI Second Brain for DevOps and Security

Your AI coding agent can now hold over one hundred and sixty specialized skills for infrastructure and security in a single library. This project acts as a massive, organized brain for tools like Claude Code and Cursor, teaching them exactly how to handle complex tasks without you having to write the instructions from scratch. It covers everything from hardening Kubernetes clusters to navigating strict compliance rules like SOC2, all packed into ready-to-run scripts and templates. The most surprising part is that it turns your everyday AI assistant into a senior-level security and operations expert by simply letting it read these structured knowledge files.

🆔 @hackernewsgithubprojects

📦 m-nabeegh/filemorrow

FileMorrow Keeps Your Downloads Private and Tidy

FileMorrow is a macOS app that organizes your Downloads folder without ever sending your files to the cloud. It uses on-device Apple Intelligence to sort documents into subject folders like Finance or Medical, while keeping everything local and private. You can undo any changes instantly, and the app safely finds duplicates and old installers to free up space. The best part is that it respects your privacy completely, meaning your files never leave your Mac. If you want a clean, safe way to manage your downloads without compromising your data, this tool is worth a look.

🆔 @hackernewsgithubprojects

📦 ransomleak/training-security-awareness

RansomLeak's 3D Office Training

RansomLeak is the open-source security training library that actually makes you learn by doing. Most corporate security courses are boring slide decks that people forget the moment they close the tab. This project flips that on its head by dropping you into a realistic, interactive thirty-dimensional office. You sit at a desk, pick up a phone, and face off against live phishing emails, scam calls, and tricky file downloads in real time. It is designed to build muscle memory, so when a real attack happens, your hands already know the right move. The entire collection is free, fully white-labeled, and ready to drop into any learning management system.

📰 https://news.ycombinator.com/item?id=48675551

🆔 @hackernewsgithubprojects

📦 gtapps/claude-code-hermit

Claude Code Hermit: The Agent That Actually Listens

Claude Code Hermit is the plugin that finally stops your coding assistant from forgetting everything between chats. Most tools reset their memory the moment you close the window, but this one stays awake, tracks its progress, and actually learns from its own mistakes. It watches for recurring issues, drafts fixes, and sends you a simple yes-or-no prompt via your phone so you stay in control without micromanaging. The best part is that it only uses energy when something actually happens, keeping your costs down while it works quietly in the background.

📰 https://news.ycombinator.com/item?id=49394794

🆔 @hackernewsgithubprojects

📦 local-inference-lab/llm-inference-bench

LLM Inference Bench: The Live Dashboard for AI Speed

LLM Inference Bench turns chaotic server logs into a live, breathing dashboard that shows exactly how fast your AI models actually generate text. Instead of guessing if your setup is fast enough, this tool tests your system across every possible combination of user load and prompt length, revealing precisely where performance breaks down. It automatically detects your backend and displays real-time GPU temperatures alongside token speeds, so you can see the physical hardware working in sync with the data. You even get instant accuracy checks against standard datasets to prove your model hasn't lost its brain during quantization.

🆔 @hackernewsgithubprojects

📦 vorssaint/vorssaint-utils

One Icon, A Dozen Mac Utilities

Vorssaint packs a dozen paid Mac utilities into a single free menu bar icon that runs entirely locally. You can install only the features you actually use, like a per-app volume mixer or window snapping, while the rest stays completely dormant. It’s privacy-focused, meaning no accounts, no telemetry, and no cloud dependencies. The app even tells you exactly which permissions each tool needs, so you keep total control over your machine. It’s the ultimate tidy-up kit for your Mac, all in one place.

🆔 @hackernewsgithubprojects

📦 lum1104/dsh-browser

dsh Browser Control: The AI That Clicks For You

Instead of using a headless bot, DSH Browser Control plugs an AI directly into your actual Chrome window. You can tell the model to fill out forms, click buttons, and read pages, and it does so using the tab you are already logged into. The interesting part is that the AI does not see the screen as an image. It reads a structured text version of the page and interacts with numbered buttons and fields. This means it works faster and protects your passwords by masking sensitive inputs before they leave your browser.

🆔 @hackernewsgithubprojects

📦 aravpanwar/decayfmt

decayfmt: Files That Dye When You Open Them

Ever watch a file rot away? No, not metaphorically. decayfmt is a Rust tool that permanently corrupts your images and text every single time you open them. You set how fast it decays by typing a number into the filename, like three for a slow fade or ten for instant static. It’s not encryption; it’s a social contract. The moment you view it, the damage is written to the drive forever, with no undo button. If you want the original, you better have a backup. It’s a weird, fascinating reminder that digital things don’t have to be immortal.

📰 https://news.ycombinator.com/item?id=49390206

🆔 @hackernewsgithubprojects

📦 notactuallytreyanastasio/shoehorn

Shoehorn: The AI Model That Fits Perfectly

Shoehorn is the tool that squeezes any large AI model into your computer's memory without wrecking its quality. Instead of guessing a standard compression setting that leaves wasted space or crashes your machine, it calculates exactly how much room you have. It then smartly decides which parts of the model deserve high precision and which can be compressed harder. The result is a file that fits perfectly into your hardware, letting you run massive models locally that would normally be impossible to host. You finally get the best performance your specific machine can handle, with zero guesswork.

📰 https://news.ycombinator.com/item?id=49299386

🆔 @hackernewsgithubprojects

📦 mengto/threeui

ThreeUI: The Open-Source 3D Component Catalog

ThreeUI is a fully open-source catalog of live, interactive 3D web components built with Three.js, and it runs entirely in your browser without requiring an account or login. It provides over fifty parent components and more than one hundred and sixty browseable results, including complex shaders, animated backgrounds, and interactive UI elements like liquid metal buttons and particle fields. The project’s unique value is its transparency; every single component comes with its complete, readable source code and assets, allowing you to copy, paste, and modify exactly what you see.

🆔 @hackernewsgithubprojects

📦 coleam00/skills

The PIV Loop: Cole's AI Coding Skills

Skills is the collection of thirty-three plain text instructions that turns coding agents into reliable software partners by enforcing a strict plan, implement, and validate workflow. Instead of relying on massive, static configuration files that confuse the model, this repository uses modular skills that load only when needed to keep context clean. The standout feature is the ability to run parallel investigations in separate git worktrees, allowing agents to diagnose issues and build features without ever touching your current working tree. It includes a meta skill that literally runs experiments to prove which of your custom rules actually matter and which are just dead weight.

🆔 @hackernewsgithubprojects

📦 wangeai/k3flight

Running a 2.8 Trillion Parameter AI on CPU

K3 Flight lets you run a massive two-point-eight-trillion-parameter AI model locally on your CPU, without needing a GPU. The model file is nearly a terabyte, but this tool uses a smart flight plan to load only the tiny piece of the brain needed for each word, keeping active memory around fifty-five gigabytes. It turns a hardware impossibility into a workable local server, proving that even the largest models can run on your desktop if you manage the data flow correctly.

🆔 @hackernewsgithubprojects

📦 fsmargoo/dsh-at-file

Point Your AI at Any File

Stop typing out long file paths for your AI. This plugin lets you type an ampersand in your prompt to search your entire project folder with one click. It acts like a smart autocomplete that finds any document, code file, or asset you need. The coolest part is that it doesn't stuff the file content into your chat. It just tags the file's location, letting the AI agent read it only when it actually needs to. This keeps your prompts clean and your context window tidy. It’s a simple, clever way to keep your AI grounded in your real work without the mess of copy-paste.

🆔 @hackernewsgithubprojects

📦 vivekhaldar/seed

The AI Agent That Builds Itself

A tiny AI agent starts life with just one tool: the ability to run shell commands. That is the entire starting kit for Seed. It does not come with pre-made skills, memory, or codebases. Instead, it grows its own brain. It writes notes, creates tools, and saves them in a local folder that acts as its persistent mind. Because it starts so small, you can see exactly how it thinks and changes. It is not a rigid framework; it is a living system that rewrites its own instructions after every chat.

📰 https://news.ycombinator.com/item?id=49384113

🆔 @hackernewsgithubprojects

📦 drstranded/co-rl

Co-RL: How AI Models Teach Each Other To Reason Without Labels

Train language models to reason by letting them grade each other's work without using any human labels. Co-RL, or Collaborative Reinforcement Learning, solves the problem of AI models getting stuck in bad loops when they only learn from their own feedback. Instead of training alone, this method uses a diverse group of different AI models to act as a team. Each model learns by looking at the answers of its peers, which breaks the cycle of repeating mistakes. This simple shift allows the system to figure out complex logic on its own. The result is a smarter model that improves without needing expensive, hand-made data.

🆔 @hackernewsgithubprojects

📦 embeddings-benchmark/embedders-dilemma

The Embedder's Dilemma: Why Cheap Beats Smart

This open-source benchmark puts ten large language models and twenty-six embedding models on the same stage for thirty-seven text tasks, but it stops there. It doesn't just measure accuracy; it tracks every dollar spent. The surprising result? While the biggest language models often win on complex reasoning, they are orders of magnitude more expensive than simple embedding pipelines. In fact, some tiny embedding models deliver nearly identical performance for mere cents. The repository provides the raw code, data, and scripts to reproduce every single chart from the paper without needing a GPU or an API key.

🆔 @hackernewsgithubprojects

📦 leonxlnx/unlazy

unlazy

Force your coding assistant to actually finish the job instead of cutting corners. AI agents have a nasty habit of stopping at eighty percent and confidently claiming victory while the code is still broken. Unlazy fixes this by making completion a strict, mechanical requirement. Before any work starts, the agent writes specific acceptance criteria into a file. Then, it has to run actual commands to prove each box is checked. It cannot just say it is done; it has to show the evidence. If the checks fail, the agent is blocked from ending the task until the work is truly complete.

🆔 @hackernewsgithubprojects