1. pdf-inspector by firecrawl - Fast Rust library for PDF inspection, classification, and text extraction.
Language: Rust · Stars: 13878 · +8641 this week · View on GitHub →
What it is. pdf-inspector is a pure Rust library that classifies PDFs as text-based or scanned, extracts position-aware text, and converts it to clean Markdown without OCR, with bindings for Python, Node.js, and browser WebAssembly.
Why it's trending. It gained 8,641 stars this week, propelled by benchmark results showing the highest overall score and fastest speed among local PDF parsers, processing text-based PDFs in under 200ms while skipping costly OCR services.
Who should care. Developers building document-processing pipelines for reports, research papers, financial documents, invoices, or legal PDFs who need fast, structured Markdown output without OCR latency or external infrastructure.
2. reverse-skill by zhaoxuya520 - Reverse Engineering / Authorized Penetration Testing / Security Research Skill Router Pack
Language: PowerShell · Stars: 22519 · +9784 this week · View on GitHub →
What it is. reverse-skill is a structured skill router pack that directs AI coding agents to the appropriate reverse engineering, authorized penetration testing, and security research methodologies and toolchains across many task types.
Why it's trending. The repository saw significant momentum this week with 9,784 new stars on its way to 22,519 total, driven by its breadth of 41 routing rules, a 163-case regression benchmark, 42 tracked skill modules, and cross-platform CI covering scenarios from APK and binary reverse engineering to CTF, malware, firmware, and red-team orchestration.
Who should care. Security researchers, reverse engineers, and authorized pentesting or CTF practitioners who use AI coding clients such as Claude Code, Codex, Cursor, OpenCode, Kiro, or Cline and want a repeatable, tool-aware workflow instead of ad-hoc command guessing.
3. TencentDB-Agent-Memory by TencentCloud - TencentDB Agent Memory is a team-level memory hub for AI Agents
Language: TypeScript · Stars: 18769 · +8003 this week · View on GitHub →
What it is. TencentDB Agent Memory is a TypeScript-based team-level memory hub for AI agents that turns conversations, documents, and code into four reusable assets—Chat Memory, Skill, Wiki, and CodeGraph—shared across agents and frameworks.
Why it's trending. It gained over 8,000 stars in a single week as its Team Memory Beta evolves quickly, with the new v3.0 release introducing a migration path from v2.x/v0.x and a one-command deployment of the memory-core, memory-hub, and proxy services.
Who should care. AI agent developers and small teams running multi-agent workflows who need persistent, portable memory and skill libraries that survive across sessions and agent frameworks.
4. airllm by lyogavin - AirLLM 70B inference with single 4GB GPU
Language: Jupyter Notebook · Stars: 30375 · +5129 this week · View on GitHub →
What it is. AirLLM is a Python library that dramatically reduces inference memory usage, enabling large language models such as 70B parameter models to run on a single 4GB GPU without quantization, distillation, or pruning.
Why it's trending. Recent updates added support for the 2.8T parameter Kimi K3 model running on under 4GB of VRAM and a v3.0 release with FP8 support for models like DeepSeek-V3 (671B) on approximately 12GB, coinciding with over 5,100 new stars this week.
Who should care. Developers and researchers who want to run very large open-source language models locally on memory-constrained hardware, including consumer GPUs and macOS systems.
5. DeepSeek-Reasonix by esengine - DeepSeek-native AI coding agent for your terminal.
Language: Go · Stars: 33450 · +4709 this week · View on GitHub →
What it is. DeepSeek-Reasonix is an open source, MIT-licensed local AI coding agent distributed as a single self-contained Go binary, engineered to run extended autonomous coding tasks with configurable safety and permission controls.
Why it's trending. It is trending this week after gaining over 4,700 new GitHub stars, as a locally run DeepSeek-native AI coding agent built for extended autonomous tasks with built-in safety controls (per-turn checkpoints, undo functionality, sandboxed workspace) and support for access via terminal, desktop app, browser, and code editor over ACP.
Who should care. Developers and engineering teams seeking a locally hosted, configurable AI coding agent for extended autonomous tasks, with customizable permissions, model choices, and multiple access interface options.
6. AI-For-Beginners by microsoft - 12 Weeks, 24 Lessons, AI for All!
Language: Jupyter Notebook · Stars: 64034 · +5514 this week · View on GitHub →
What it is. AI-For-Beginners is a 12-week, 24-lesson beginner-friendly artificial intelligence curriculum from Microsoft that includes practical lessons, quizzes, labs, and covers foundational AI concepts, TensorFlow, PyTorch, and AI ethics.
Why it's trending. It is trending this week thanks to its automated, always up-to-date translations in over 50 languages, which makes its structured beginner AI curriculum accessible to learners worldwide.
Who should care. Absolute AI beginners, educators looking for a ready-to-use introductory AI curriculum, and anyone seeking accessible, hands-on AI learning with ethics content are the core audience for this project.
7. kaneo by usekaneo - 🎯 All you need. Nothing you don't.
Language: TypeScript · Stars: 7905 · +1952 this week · View on GitHub →
What it is. Kaneo is an open-source, self-hosted minimalist project management platform built to avoid the feature bloat that distracts teams from core productive work.
Why it's trending. It is trending this week with 1,952 new GitHub stars, driven by its built-in MCP server that enables native task management integration with AI tools including Claude and Cursor, and its focus on eliminating unnecessary features common in bloated project management platforms.
Who should care. Teams, developers, and users of AI productivity tools like Claude and Cursor who want a lightweight, self-hosted project management solution that prioritizes simplicity over feature clutter.
8. book-to-skill by virgiliojr94 - Turn any technical book PDF into a Claude Code skill
Language: Python · Stars: 19442 · +4121 this week · View on GitHub →
What it is. book-to-skill is a Python tool that turns technical books and other structured prose into an Agent Skills–compatible package (a SKILL.md plus per-chapter, glossary, patterns, and cheatsheet files) that AI coding agents like Claude Code, GitHub Copilot CLI, and Amp can load on demand.
Why it's trending. It addresses the common pain of agents hallucinating or forgetting book content by structuring material once at conversion and reporting 24×–51× fewer tokens than re-feeding full books into context, a combination that helped it gain roughly 4,121 stars this week on top of 19,442 total.
Who should care. Developers, technical writers, and teams using Claude Code, GitHub Copilot CLI, or Amp who want their AI agent to reliably recall and answer from technical books, internal documentation, brand/design guides, or reference specs.
9. skills by google - Agent Skills for Google products and technologies
Language: Python · Stars: 17231 · +1626 this week · View on GitHub →
What it is. It is a repository of installable Agent Skills for Google products and technologies, with a broad catalog of Google Cloud skills spanning AI/ML, GKE infrastructure, databases, and analytics.
Why it's trending. It gained over 1,600 stars this week as an actively developed collection of Google Cloud agent skills covering model deployment, GKE management, and analytics.
Who should care. Developers and cloud architects working with Google Cloud who want ready-to-use agent skills for AI/ML, Kubernetes, databases, and analytics workflows.
10. swarm-forge by unclebob - A simple tool for coordinating several AI agents.
Language: Clojure · Stars: 2052 · +562 this week · View on GitHub →
What it is. SwarmForge is a tmux-based agent orchestration platform that coordinates swarms of AI agents working in separate git worktrees via shared scripts and role-specific prompts.
Why it's trending. The repository gained 562 stars this week (reaching 2,052 total), reflecting strong interest in its structured multi-agent coding workflows built around role-separated agents such as coder, cleaner, refactorer, architect, hardender, and QA.
Who should care. Developers and teams experimenting with multi-agent AI coding pipelines using backends like codex, claude, copilot, or grok who want disciplined, role-based collaboration across tmux sessions and git worktrees.