📦 msnightmare/falconflank

FalconFlank: The Office Macro Bypass

FalconFlank is the security research project that exposes a hidden gap in CrowdStrike’s endpoint protection. It proves that the software meant to automatically strip malicious macros from office files actually leaves a tiny opening for attackers to slip through and grab full control of a computer. The code shows how a standard office document can be used to run dangerous code that the sensor completely misses, even on the latest versions of Windows. It is a real example of how well-intentioned security features can accidentally create new risks. If you care about digital safety, remember that no system is perfect and testing these weaknesses helps us all stay safer.

🆔 @hackernewsgithubprojects

📦 davidmosiah/google-health-mcp

Google Health MCP: Your AI Agent's Private Health Access

Google Health MCP lets your AI chatbot read your actual Fitbit or Pixel Watch data right on your own computer, without ever sending your private login tokens to the cloud. Most health apps keep your stats locked away, so this tool acts like a private bridge that lets your local assistant check your sleep, heart rate, and daily movement safely. It works by running a secure, local connection that strips away any identifying details before your AI even sees the numbers. The coolest part is that you get personalized insights without sacrificing your privacy to a third-party server.

🆔 @hackernewsgithubprojects

📦 tstanczyk95/mcbyteplusplus

McByte++: Tracking Sports Without Training

McByte++ is the tracking tool that follows athletes without ever needing to train on their specific game. This project is the official code for a research paper that solves a huge headache in sports video analytics. Normally, you have to spend ages teaching a computer who is who, but this method just works right out of the box. It keeps track of every player on the court, and if they leave the frame and come back, it remembers exactly who they are. It is built by researchers at Inria and offers a smarter, faster way to analyze sports video without the usual heavy lifting.

🆔 @hackernewsgithubprojects

📦 lmqfly/agentfold

AgentFold: AI-Driven Protein Folding Discovery

A computer program now independently designs better protein folding models by writing its own code, fixing its own bugs, and training its own variants. AgentFold is a multi-agent framework that automates the entire scientific discovery loop for protein structure prediction. It starts with an existing model, proposes specific architectural changes, implements the code, runs the training, and evaluates the results using structural metrics. The system then analyzes what worked and what failed, storing these lessons to guide its next attempt. This creates a self-improving cycle where the AI refines the model based on real performance data, not just guesswork.

🆔 @hackernewsgithubprojects

📦 funinkina/openeffects

OpenEffects: Linux Webcam Effects

OpenEffects is the Linux webcam tool that finally puts professional camera tricks right on your desktop. It acts like a smart filter for your webcam, letting you blur your messy room, swap in a new background, or let the camera automatically zoom in on your face so you always look centered and professional. The coolest part is that it runs entirely on your computer, so your video calls stay private and snappy without needing a separate app to run in the background. It works with any program you use, meaning you get these fancy effects without ever leaving your meeting or streaming software.

📰 https://news.ycombinator.com/item?id=49547949

🆔 @hackernewsgithubprojects

📦 opentslm/timerlm

How TimeRLM Finds Tiny Errors in Giant Data

Spotting a glitch in a million data points sounds impossible, but TimeRLM makes it routine by letting an A.I. explore your data like a detective instead of just staring at a chart. Instead of squishing a huge time series into a tiny summary, the model opens a live code window to zoom in, run calculations, and hunt down the exact moment something broke. This is brilliant because it catches minute errors that would vanish in a standard snapshot, turning a massive, messy signal into a solvable puzzle. It is a clever way to give A.I. real investigative tools rather than just raw numbers.

🆔 @hackernewsgithubprojects

📦 lanrhyme/micyou

MicYou: Use Your Phone as a PC Microphone

MicYou turns your Android phone into a high-quality microphone for your computer by streaming audio over Wi-Fi or USB. It solves the problem of poor built-in PC mics by using your phone's superior hardware for calls and recording. The software routes this audio through a virtual device so any application can use it as if it were a physical plug-in microphone. It even includes AI noise suppression to clean up the sound. This is a simple way to get broadcast-quality audio without buying new equipment. It is a great tool for streamers and anyone needing better voice input.

🆔 @hackernewsgithubprojects

📦 qufei1993/skills-hub

Skills Hub: One App for All AI Coding Tools

A single desktop application now manages skills for over forty different AI coding tools simultaneously. This solves the messy problem of keeping instruction files updated across Cursor, Claude Code, and other assistants, where manual copying leads to version conflicts and wasted effort. The app installs every skill into one central folder, then automatically links them to whichever tools you have active, so you update a capability in one place and see it appear everywhere instantly. This centralized approach removes the need to track down every hidden configuration folder on your computer. It is the definitive way to organize your AI toolkit without the clutter.

🆔 @hackernewsgithubprojects

📦 shopify/claude-for-commerce-examples

Claude for Commerce Examples: The AI That Shops and Sells

Claude for Commerce Examples is the practical blueprint that finally makes AI shopping agents work on real Shopify stores. It connects a customer-facing chatbot to your live catalog, letting buyers search, build carts, and ask policy questions without leaving the page. The real magic is in the merchant side, where the AI reads your inventory and proposes price changes or restocks, but only applies them after you explicitly hit approve. It is the perfect starting kit if you want to build custom AI agents that actually touch your store data, giving you full control over what gets changed and when.

🆔 @hackernewsgithubprojects

📦 databricks-industry-solutions/lakehouse-industry-data-models

Forty Industries, One Click

An AI agent just generated forty full data models for different industries, complete with tables and foreign keys. This repo gives you a working blueprint for any business, saving you from drawing database diagrams from scratch. It even fills those tables with realistic sample data, so you can test your code immediately without hunting for real files. Pick an industry like banking or restaurants, and the installer sets up your entire database structure in one go. It is the fastest way to build a working data demo without weeks of manual design. Grab one, install it, and see your data flowing in minutes.

📰 https://news.ycombinator.com/item?id=49549309

🆔 @hackernewsgithubprojects

📦 vedaant00/opendot

The AI Agent You Can Actually Undo

Most terminal AI agents will happily delete a file or run a dangerous command, leaving you to clean up the mess. OpenDot changes the rules by taking a full snapshot of your workspace before it does anything. If the agent messes up, you simply type undo, and your files snap back exactly as they were, down to the byte. It works with any AI model you like, from OpenAI to local runs, and it even warns you before doing things that cannot be reversed. It is like having an AI assistant with a safety net, so you can let it work without holding your breath.

📰 https://news.ycombinator.com/item?id=49100984

🆔 @hackernewsgithubprojects

📦 orron/epicinfographics

Epic Infographics: The AI Skill That Kills Generic Charts

Epic Infographics is the open source skill that stops AI from making those boring, generic charts. You know the ones with the blue boxes and random emojis. This tool teaches your AI agent to design data the way a professional studio would. It builds real scenes, like a blueprint or a magazine spread, instead of flat pages. It even animates the final graphic for you. If you are tired of your AI looking like a robot, this is the fix. Go install it and make something that actually looks cool.

🆔 @hackernewsgithubprojects

📦 rakanki911/dlss5-swapper

DLSS 5 Swapper: One Click Upgrade

Install DLSS 5 Neural Rendering into your entire game library with a single click. DLSS 5 Swapper is a desktop tool that scans your Steam, Epic, and GOG folders to find compatible titles, then automatically installs the AI upscaling files for you. The real magic is its Feeder mode, which lets you add modern AI graphics to older games that lack native support, effectively future-proofing your entire catalog. It even handles emulators, so you can run retro games with the latest visual tech. If you want your games to look crisp without digging through obscure forums, this tool is your shortcut. It just works.

🆔 @hackernewsgithubprojects

📦 darama22/malware-research-hub

Malware Research Hub: 2,700 Real Samples in One Place

Explore fifty-three years of computer history through two thousand seven hundred and sixty-four real malware specimens. Malware Research Hub bundles every major virus family, from the first Creeper worm in nineteen seventy-one to modern ransomware, into one searchable local application. The project solves the chaos of scattered digital archives by indexing encrypted samples and organizing them into a clean, bilingual interface. You can filter by platform or type to see exactly how these threats evolved without touching the actual code. This local tool gives researchers a safe, structured way to study the anatomy of cybercrime. Keep it running offline to explore the timeline of digital danger at your own pace.

🆔 @hackernewsgithubprojects

📦 daniilsokolyuk/video2dlssnr

Video2dlssnr: The Secret to Game-Grade Video Enhancement

Video2dlssnr is the Windows tool that finally lets you run real-time game graphics magic on your own video files. It takes your raw footage and pushes it through the exact same neural upscaling engine found in modern video games, adding crisp, high-definition detail that standard software just cannot touch. You get a simple command-line utility or a friendly web interface to process clips, keeping every frame on your graphics card for lightning-fast speeds without sacrificing quality. It is the perfect way to take a rough recording and give it a professional, cinematic finish, proving that you do not need a studio to get stunning results.

🆔 @hackernewsgithubprojects

📦 cfal/garcon

Garcon: Run Coding Agents in Your Browser

Garcon lets you run seven different coding agents side by side in a single browser window on your own machine. The biggest surprise is that you can literally message one agent to another, letting them hand off tasks or review each other's work in real time. It turns a messy terminal workflow into a tidy, tiled workspace where you can steer active tasks, review code changes, and ship updates without ever opening a separate app. If you are juggling multiple AI helpers and losing track of who is doing what, this is the control center that keeps them all in check.

🆔 @hackernewsgithubprojects

📦 coldteadotai/pr-lens

PR Lens: Visualize Code Changes Instantly

PR Lens is the code review tool that shows you the whole picture before you read a single line. Instead of scrolling through hundreds of lines of code, it automatically draws an animated diagram right inside your pull request. This tool maps out exactly which parts of your system are changing, using colors to show what is new, changed, or removed. You can even watch the data flow through these steps in real time. This is huge because it turns a confusing wall of text into a clear visual map, helping you understand the impact of your changes instantly.

📰 https://news.ycombinator.com/item?id=49466095

🆔 @hackernewsgithubprojects

📦 hardbeat920/monocode

MonoCode: One App for All Your Coding Agents

Run every major coding agent from one desktop window. MonoCode sits on top of the command-line tools you already have, like Claude Code, Codex, and Cursor, and gives them a real interface. Instead of juggling separate terminals, you get tabs for your sessions, a built-in editor, and a live terminal right in front of you. The standout feature is the handoff card. When an agent finishes a task, you can instantly pass that context to a different provider in a split pane without losing your place. It is a lightweight layer that turns scattered command-line workflows into a single, manageable workspace.

🆔 @hackernewsgithubprojects

📦 bybit-exchange/svg-diagram

Svg-diagram: The House Style for Perfect Diagrams

Turn raw code into polished architecture diagrams that actually look professional. This is svg-diagram, a new tool that hands your AI agent a strict rulebook for drawing charts. Instead of letting the model guess where boxes go, it forces every line, curve, and label to follow one exact visual style. The coolest part is the built-in linter, a zero-dependency checker that scans your finished file and tells you exactly what is wrong, from bad spacing to missing fonts. It turns messy AI output into clean, readable art you can trust. No more squinting at crooked arrows; just ask for a diagram, run the check, and ship it.

🆔 @hackernewsgithubprojects

📦 royalbhati/sqltoerdiagram

Turn SQL into a Diagram Instantly

Turn raw database code into a clear picture instantly. This browser tool takes your table definitions and draws them out as a map, connecting tables that relate to each other. It solves the problem of squinting at complex text to understand your data structure. Everything runs locally in your browser, so your sensitive database layout never leaves your computer. You can drag tables around, edit names directly on the screen, and even share the entire diagram with a single link. It is a great way to visualize your database without uploading anything. Stop guessing your schema relationships and see your data structure clearly.

🆔 @hackernewsgithubprojects

📦 junchao-cs/solarwm

SolarWM: The One-Clip Trick for Endless Video

SolarWM is the video training framework that turns tiny five-second clips into endless, interactive worlds. It solves the problem of short memory by using a clever three-step recipe that teaches models to predict the future from just a brief glimpse. The surprising part is that these models can roll out for hours without losing the plot, all from training on mere seconds of footage. It’s a fully open foundation that handles everything from data cleanup to final generation, making it the go-to for building interactive video experiences that actually last. If you want to create video worlds that feel continuous and responsive, this is the toolkit you need to explore.

🆔 @hackernewsgithubprojects