56 subscribers
7.29K videos
7.9K links
Download Telegram
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ saint-viperx/scr_bench

Is Your AI Agent Safe? Meet SCR-Bench

Are your AI agents truly secure? While individual skills might seem safe, SCR-Bench reveals they can become dangerous when combined. This framework evaluates critical risks that emerge only during interaction, focusing on how benign skills create harmful paths through capability flow, trust transfer, and authorization confusion. With sub-benchmarks like CapFlow and TrustLift, it exposes vulnerabilities that isolated reviews completely miss, showing dramatic spikes in attack success rates. By testing these compositions, developers can identify hidden security flaws in agent workflows before they cause real-world damage. Explore this essential benchmark to better secure your AI agent ecosystems today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ cursor/agent-trace

Tracking AI Code Contributions

Ever wonder if your code was written by you or an AI agent. As AI coding tools grow in popularity, keeping track of who authored what has become a major challenge for developers. Agent Trace solves this by providing an open, vendor-neutral standard for recording attribution in your codebase. It defines a consistent way to log whether specific lines were created by humans, AI, or a collaboration of both. This specification makes it easy to track AI contributions alongside your own work, ensuring transparency without locking you into any specific tool. Start tracing your agent interactions today.

πŸ“° https://news.ycombinator.com/item?id=48616614

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ tanlethanh/zedra

Zedra: Control Your AI Coding Agents from Your Phone

Ever wish you could manage your AI coding agents even when you are away from your desk? Meet Zedra, an experimental mobile control plane that bridges the gap between your smartphone and your desktop workspace. By pairing a lightweight desktop daemon with a high-performance mobile app, it lets you monitor, steer, and interact with agents like Claude Code or Codex directly from your phone. Built with Rust and GPU-accelerated rendering, it provides a smooth, secure experience using direct peer-to-peer connections. Take control of your agentic workflows and stay productive from anywhere by giving your mobile device a real seat at the table.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ th0rgal/open-ralph-wiggum

Automate Your Coding with Open Ralph Wiggum

Are you tired of AI agents struggling to complete complex tasks in one go? Meet Open Ralph Wiggum, a powerful tool that wraps your favorite AI coding agents into an autonomous, self-correcting loop. Whether you use Claude Code, Copilot, or others, this CLI tool repeatedly runs your prompt, allowing the agent to see its own previous work and learn from mistakes until the job is done. It supports features like task tracking and mid-loop guidance, helping you build full applications without constant babysitting. Simply set your goal, define your iterations, and let the loop handle the rest for a truly productive workflow.

πŸ“° https://news.ycombinator.com/item?id=48616864

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ unmodeled-tyler/vessel-browser

Vessel Browser: The AI-First Web Browser for Autonomous Agents

Tired of AI agents operating in headless, invisible silos? Meet Vessel Browser, the open-source, Chromium-based browser designed specifically for autonomous agents. Instead of running in a black box, Vessel provides a persistent, human-visible interface where agents drive and humans stay in the loop to audit or steer workflows. With features like durable sessions, structured page schema inference, and a supervisor sidebar for real-time tracking, it transforms browser automation into a transparent, collaborative experience. Whether you are using Hermes Agent or your own custom API provider, Vessel offers the control and visibility needed to make long-running web tasks truly reliable.

πŸ“° https://news.ycombinator.com/item?id=48576073

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ sumitgautam0101/windows-copilot-api

Use Microsoft Copilot as a Free API

Want to build your own apps with Microsoft Copilot without paying for expensive API keys or billing plans? This project is a total game changer because it reverse engineers the consumer Copilot experience into a fully functional, OpenAI-compatible API. By logging in once with your personal account, you can access powerful AI models directly through Python or by running a local server. It even supports streaming and multi-turn conversations, allowing your code to interact with Copilot just like a standard OpenAI client. It is the perfect way to integrate advanced AI into your own projects completely for free.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ thousandflowers/raccoon

Take Control of Your Mac with Raccoon

Are you a macOS power user tired of juggling multiple apps just to keep your system running smoothly? Meet Raccoon, a lightweight, Bash-based toolkit designed to put essential system diagnostics and security audits directly into your terminal. This powerful companion helps you monitor everything from battery health and disk status to complex security scans and developer workflows like updating Homebrew and npm packages all at once. With its optional terminal user interface, you get a clean, visual way to manage your Mac effectively. It is a must-have utility for anyone wanting to maintain a lean, secure, and organized system.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ agavra/tuicr

Master Your Code Reviews with This Terminal Tool

Stop struggling with slow web interfaces for code reviews and try a faster, keyboard-driven workflow. This powerful terminal user interface brings professional code review capabilities directly into your command line. It features seamless vim-style navigation, allowing you to scroll through code changes, add line or file comments, and track your review progress with ease. Whether you are using git, mercurial, or jujutsu, it handles your diffs in one continuous stream. Once finished, you can push your feedback directly to platforms like GitHub or GitLab or export structured markdown for your favorite coding assistant to polish your work.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ ahekot/comfyui_vnccs_utils

Supercharge ComfyUI with VNCCS Utils

Unlock a more powerful creative workflow in ComfyUI with this impressive collection of utility nodes. This repository introduces the UniCanvas, an integrated infinite-canvas workspace for seamless layer-based image editing and generation, alongside the Pose Studio, which provides a professional 3D environment for posing characters with interactive lighting and camera controls. Whether you need precise mask tools, automated model management for your favorite loaders, or advanced detailer nodes, these utilities streamline your complex projects without leaving the graph. Boost your efficiency and take full control of your creative process by integrating these specialized tools into your everyday workflows today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ kyuz0/amd-r9700-vllm-toolboxes

Supercharge Your AMD R9700 for Local AI!

Struggling to get top-tier AI performance on your AMD Radeon R9700? This repository is the ultimate solution for running local LLMs with ease. It provides specialized, containerized Toolbx environments specifically tuned for the gfx1201 architecture, letting you bypass complex dependency issues while keeping your host system clean. By integrating the AITER Unified Attention backend, these containers resolve performance bottlenecks, particularly during long-context inference where standard setups often fail. Whether you need an interactive development environment or a robust deployment service, these toolboxes make high-performance AI inference on AMD hardware finally feel seamless. Give your setup the boost it deserves today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ ai-infos/vllm-gfx906-mobydick

Run LLMs on Older AMD GPUs: GFX906 Optimization Guide

Do you have an older AMD GPU sitting around? This project brings high-throughput LLM serving to hardware that might otherwise be left behind. By providing a specialized fork of the popular vLLM engine, this repository enables optimized inference for AMD gfx906 cards like the Radeon VII, MI50, and MI60. It solves compatibility hurdles by including custom builds for ROCm, Triton, and FlashAttention tailored to these specific architectures. Whether you choose the provided pre-built Docker container or follow the manual build steps, you can finally run modern large language models efficiently on your existing hardware. Unlock your GPU's potential today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ lisj575/omnidirector

OmniDirector: Bring Your Images to Life with Professional Camera Moves

Have you ever wanted to animate static images with complex camera movements? Meet OmniDirector, an innovative tool designed to clone diverse camera motions from multi-shot videos directly onto your source images. By utilizing a unique camera grid representation, this system solves the challenge of animating footage without needing cross-paired data. It even features a smart hierarchical prompt expansion agent that balances multiple control signals for seamless results. Whether you are a creator looking to level up your visual storytelling or just curious about the future of motion synthesis, this project offers a powerful way to master cinematic control.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ zarazhangrui/codebase-to-course

Turn Any Codebase Into An Interactive Course

Ever wanted to understand how a codebase really works without spending years on a computer science degree? This clever tool acts as a dedicated guide, transforming any repository into a beautiful, self-contained interactive course. It is designed specifically for vibe coders who want to master their projects by tracing real-world data flows and logic through animated visualizations, quizzes, and side-by-side plain English code translations. By turning abstract code into a visual learning journey, it empowers you to debug faster and steer AI tools with total confidence. Start building your own interactive walkthroughs today and master your tech stack.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ 2arons/duel-agents

Slash Your AI Costs with Duel Agents

Are you tired of overpaying for AI coding agents that use expensive flagship models for every single task? Duel Agents is an ingenious IDE-native routing layer designed to help you save significantly on AI bills without sacrificing quality. It works by running your prompts against multiple models simultaneously and intelligently selecting the cheapest response that still meets your performance standards. This repository provides the official integration tools, including a CLI for easy setup, an SDK for custom development, and seamless plugins for tools like Cursor and Claude Code. Stop burning your budget and start routing smarter with Duel Agents today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ tashfeenahmed/freellmapi

Get 1.7 Billion Free AI Tokens with One Endpoint

Are you tired of managing dozens of different free AI model accounts and struggling with various rate limits? This open-source tool is a game changer, acting as an intelligent proxy that aggregates the free tiers of sixteen major LLM providers into a single, seamless endpoint. It handles automatic failover if a provider hits a limit, tracks your usage to keep you within free caps, and even encrypts your API keys for security. Simply point your favorite AI apps at this local server to access a massive library of models with one unified key. It is the smartest way to experiment freely.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ jumbocontext/jumbo.cli

Take Control of Your Coding AI with Jumbo

Are you struggling to keep your AI coding assistants on track? Meet the command line tool designed to master context and memory. This tool solves the frustration of scattered project knowledge by letting you explicitly manage goals, architectural decisions, and technical guidelines. It uses a structured event-driven system to track your project components, dependencies, and evolving requirements, ensuring your coding agents stay perfectly aligned with your specific needs. Stop wasting time re-explaining the big picture to your AI and start orchestrating your development process with precision. Give your coding workflow the context it actually deserves.

πŸ†” @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
πŸ“¦ neuralinitiative/claude-dnd-skill

Turn Claude into Your Ultimate D&D Dungeon Master

Ever wanted a Dungeon Master that is available anytime, perfectly tracks 5e rules, and brings your campaign to life on your TV? This repository turns Claude Code into a powerful, persistent D&D engine. It handles everything from rolling dice and tracking character stats to voicing NPCs and managing complex, multi-session campaigns. You can even use the cinematic display companion to stream typewriter-style narration and live stats to any screen in your room while players submit actions from their phones. It is the perfect way to run a seamless, immersive tabletop experience without needing an extra person to manage the table.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ halfof8/patterngen-oss

Automate Motion Graphics with PatternGen

Struggling to add professional flair to your motion graphics? Meet PatternGen, a creative toolchain designed to streamline your animation workflow. It allows you to design title compositions in Figma, then brings them to life in the browser with procedural, deterministic pattern fills and smooth animations. You can generate layered sequences and export them as transparent PNGs for seamless integration into Blender. By using seeded random placement, it ensures consistent results every time. It is a fantastic way to add complex, rhythmic visual elements to your projects with ease. Level up your design process and start animating today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ ndvbd/niv

AI Magic: Turn Any Font into a Variable Font!

Ever wanted to turn a static font into a flexible variable font? Neural Axis Variations makes it possible. This project provides a powerful framework for generating variable font axes like weight, width, and slant using machine learning. By processing TTF files into structured training data, it allows you to train models that intelligently adapt font properties. You can easily download existing datasets or prepare your own to create custom variable outputs from single fonts. It is a brilliant solution for designers and developers looking to modernize typography workflows. Dive in and start transforming your type collection into something truly dynamic today!

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ youmind-openlab/awesome-nano-banana-pro-prompts

Unlock Google Nano Banana Pro with This Giant Prompt Library

Struggling to get the best results from Google's Nano Banana Pro AI? This repository is your ultimate shortcut to mastering high-quality image generation. It serves as the world’s largest curated prompt library, housing over ten thousand expert-verified prompts across seventeen languages. Instead of guessing, you can explore creative compositions and lighting styles that actually work. With automated GitHub workflows, the collection stays fresh and updated every four hours. Whether you are a beginner or a pro, start experimenting with these community-vetted prompts today to unlock the full artistic potential of your AI models and create something truly stunning.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ zhuoyangliu2005/t-rex

T-Rex: Leveling Up Robot Dexterity with Tactile Feedback

Ever wonder how robots can handle delicate objects without crushing them? Most AI models rely solely on vision, but T-Rex changes the game by introducing tactile-reactive dexterous manipulation. This project tackles the limitation of static touch sensors by using a high-frequency, asynchronous Mixture-of-Transformers architecture. It processes touch signals at twenty hertz, allowing for real-time force refinement without slowing down the primary vision stack. By training on a massive dataset of motor primitives, the system achieves over thirty percent higher success rates in complex tasks. If you are into robotics, check out this open-source framework to elevate your agent's touch capabilities.

πŸ†” @hackernewsgithubprojects