Cool Projects on GitHub

Claude Code skill that cuts token usage by 65% by making the AI speak in compressed caveman shorthand — same results, far fewer tokens.

A massive crowd-sourced collection of leaked system prompts from ChatGPT, Claude, Gemini, Grok, Cursor, and dozens more AI products.

Persistent memory across sessions for any AI agent — captures, compresses, and injects context back into Claude Code, Codex, Gemini, Copilot, and more.

Sesame AI Labs' Conversational Speech Model — ultra-realistic voice generation with natural prosody and emotion.

Fast, lightweight open-source TTS model (82M params) — high-quality voice synthesis that runs locally with minimal compute.

Microsoft's open-source Python tool that converts any file or Office document — PDFs, Word, Excel, PowerPoint, images — into clean Markdown.

Fair-code workflow automation platform with native AI capabilities — 400+ integrations, visual builder, custom code, and self-host or cloud.

Self-hosted AI workspace — a private, open-source alternative to commercial AI platforms you fully control.

Human-sounding open-source TTS with emotional expression — real-time capable, LLM-based speech synthesis by Canopy AI.

The leading agent meta-harness for Claude — deploy multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems.

Gives your AI agent good taste — stops it from generating boring, generic slop by injecting aesthetic judgment into every output.

Zyphra's open-weight TTS model trained on 200k+ hours — expressive multilingual speech matching top commercial providers.

In-depth tutorials and notebooks on LLMs, RAG pipelines, and real-world AI agent applications.

817 structured cybersecurity skills for AI agents, mapped to MITRE ATT&CK, NIST CSF 2.0, D3FEND, and more — works with Claude Code, Cursor, Copilot, and 20+ platforms.

A curated collection of 70+ DESIGN.md files reverse-engineered from real websites — drop one into your project and let AI coding agents generate a UI that matches the brand.

Open-source browser automation platform that lets AI agents use the internet like a human — click, type, navigate, and extract data from any website.

Open-source, LLM-friendly web crawler and scraper that outputs clean markdown and structured data built for RAG and AI pipelines.

Claude Code skill that discovers and installs the right agent skill for whatever you're working on, from the Vercel Labs skills package manager.

Claude Code/Cursor/Codex skill that turns a codebase — docs, SQL schemas, configs, PDFs included — into a queryable knowledge graph.

Claude Code skill that strips AI tells from generated text, making it read naturally as human-written prose.

Converts a reference image of an object into a procedural, animation-ready Three.js model — built from primitives, shaders, and generated geometry instead of photogrammetry.

A design skill for AI coding agents that gives them 23 precise commands and 59 detector rules to stop shipping "AI slop" UI and start producing polished, on-brand interfaces.

Low-code visual builder for AI agents and workflows — drag-and-drop LangChain-style pipelines with instant API deployment.

MoE LLM with RL-trained reasoning from InclusionAI — open-source, multilingual, built for complex reasoning tasks.

Andrej Karpathy's pattern for AI-maintained knowledge bases that compound over time — a durable alternative to re-querying RAG on every question.

A plug-and-play library of marketing skills for Claude Code and AI agents — CRO, copywriting, SEO, analytics, and growth engineering.

Open-source no-code platform that turns any website into a structured API — build web scrapers and crawlers by clicking, not coding.

Directory of 150+ free, open-source web tools that need zero signup — design, dev, media, productivity, and privacy utilities.

Family of open-weight models with hybrid Mamba-Transformer architecture and 1M-token context window. Designed for specialized AI agents with multimodal support across video, audio, image, and speech.

Self-hosted, user-friendly chat interface for local and API-based LLMs — supports Ollama, OpenAI-compatible APIs, RAG, and more.

AI software engineering agent platform that can write code, run commands, browse the web, and call APIs to complete real dev tasks.

Claude Code skill that turns a name or handle into a scored OSINT dossier — psychoprofile, career map, and confidence grades.

Unified AI agent toolkit — single LLM API, agent loop, TUI, and coding agent CLI in one TypeScript package.

Reasoning MoE LLM open-sourced by InclusionAI — Mixture-of-Experts architecture with strong reasoning capabilities.

Optimizes natural-language skill documents for AI agents without touching model weights — evolving skills through rollouts, reflection, bounded edits, and validation.

VIP cheatsheets for Stanford's CS 229 Machine Learning course — concise visual summaries of every major ML concept.

Fast reasoning model by StepFun offering high-speed inference for complex tasks. Open-source model optimized for speed without sacrificing reasoning quality.

Free, open-source PDF editor and processor — merge, split, convert, compress, OCR, and edit PDFs, self-hosted or via web app.

Autonomous pentesting platform that finds, validates, and auto-fixes vulnerabilities across APIs, web apps, code, infrastructure, and cloud — with merge-ready PRs and CI/CD gating.

End-to-end user research skill for AI agents — AI interviews, synthetic users, quant surveys, and participant recruitment in one toolkit.

Open-source AI-powered answering engine — ask questions, get cited answers, self-host for full control.

Microsoft's open-source voice AI framework — long-form, multi-speaker text-to-speech and transcription, plus a lightweight real-time streaming TTS model.