| # | Repository | ⭐ Today | Total ⭐ | Category |
|---|---|---|---|---|
| 1 | huangruiteng/loopx NEW CATEGORY | +327 | 1,740 | Agent Loop Engineering |
| 2 | TencentCloud/TencentDB-Agent-Memory | +1,891 | 14,638 | Team-Level AI Memory Hub |
| 3 | firecrawl/pdf-inspector | +1,583 | 10,818 | PDF Intelligence |
| 4 | addyosmani/agent-skills MASSIVE | +203 | 81,759 | Agentic Engineering Skills |
| 5 | uber/ADR HOT | +354 | 866 | Enterprise AI Agent Security |
| 6 | lyogavin/airllm | +833 | 28,772 | Local AI Inference |
| 7 | obra/superpowers CATEGORY-DEFINING | N/A | 266,458 | Agentic Skills Framework |
| 8 | esengine/DeepSeek-Reasonix HOT | N/A | 31,211 | DeepSeek-Native Coding Agent |
| 9 | donnemartin/system-design-primer | Popular | Popular | System Design |
huangruiteng/loopx (1,740⭐, +327/day) is birthing a brand-new category: Agent Loop Engineering. This lightweight loop engineering state kernel solves the fundamental problem of managing long-running AI agent teams. As agent teams grow from single agents to multi-agent swarms, the need for structured loop engineering — orchestration, state persistence, and lifecycle management — becomes critical. This is the first dedicated project in this space.
Addy Osmani's agent-skills has exploded to 81,759⭐, establishing a new category: Production-Grade Engineering Skills for AI Coding Agents. This is the Chrome DevTools veteran and Google engineering leader creating the playbook for how AI agents write production code. The "agentic engineering skills" category is being defined right now at enterprise scale. This is one of the biggest stories in the AI coding agent ecosystem.
uber/ADR continues its strong growth trajectory, now at 866⭐ with +354/day. Deployed at Uber, this project is the leading enterprise AI agent security framework. The $3.5B+ enterprise AI security market is taking notice. Agent-level observability, security benchmarking, and threat detection for AI agents — this is the next frontier.
Tencent's team-level memory hub continues explosive growth, adding +1,891⭐ today to reach 14,638⭐ total. This is the fastest-growing project on today's list by daily stars. The concept of turning conversations, docs, and code into reusable memory assets across agent teams is clearly resonating at enterprise scale.
firecrawl/pdf-inspector keeps climbing, now at 10,818⭐ with +1,583/day. The Rust-based PDF inspection, classification, and text extraction library has become the go-to tool for PDF intelligence. Smart routing decisions based on scanned vs text-based PDFs — this zero-competition category is maturing fast.
AirLLM continues to dominate local AI inference, now at 28,772⭐. Running 70B models on a single 4GB GPU is the killer feature. The local AI movement is accelerating, and AirLLM remains the category king.
obra/superpowers at 266K+⭐ remains the definitive agentic skills framework. The concept of subagent-driven development with agentic skills is now a well-established software development methodology. This category is maturing, and superpowers is the gold standard.
DeepSeek-Reasonix continues to grow as the premier DeepSeek-native AI coding agent. At 31,211⭐, it's proving that the DeepSeek ecosystem has enduring demand for specialized coding agents. The terminal agent category is expanding.
1. Agent Loop Engineering is born. huangruiteng/loopx (1,740⭐, +327/day) creates the Agent Loop Engineering category. Managing long-running AI agent teams needs dedicated loop engineering infrastructure.
2. Addy Osmani defines agentic engineering skills. agent-skills (81,759⭐) is a category-defining project from a Google engineering leader. Production-grade skills for AI coding agents is the new standard.
3. Enterprise AI Agent Security is real. uber/ADR (866⭐, +354/day) proves the market. Agent-level security, observability, and threat detection deployed at Uber.
4. Team-level memory goes viral. TencentDB-Agent-Memory (+1,891/day) is the fastest-growing project today. The team-level AI memory hub category is a $B+ opportunity.
5. PDF Intelligence keeps climbing. pdf-inspector (10,818⭐, +1,583/day) is the definitive PDF inspection tool in a zero-competition category.
6. Local AI inference is unstoppable. AirLLM (28,772⭐, +833/day) continues to dominate. Running 70B on a single 4GB GPU is the killer use case.
7. Weekly ecosystem is exploding. book-to-skill (+5,420/week), block/buzz (+7,262/week, new hive mind communication), and AI-For-Beginners (+8,582/week) show the ecosystem expanding into new categories like education, communication, and skill automation.
| # | Repository | Weekly ⭐ | Total ⭐ | Notable |
|---|---|---|---|---|
| 1 | microsoft/generative-ai-for-beginners | +8,582 | 61,937 | AI Education |
| 2 | zhaoxuya520/reverse-skill | +8,386 | 18,780 | Reverse Engineering Skill Router |
| 3 | block/buzz NEW | +7,262 | 22,788 | Hive Mind Communication |
| 4 | akshayballal95/book-to-skill EXPLODING | +5,420 | 16,748 | Book → Skill Automation |
| 5 | ayghri/i-have-adhd | +4,389 | 17,040 | ADHD-Friendly Agent Output |
| 6 | lyogavin/airllm | +3,911 | 28,772 | Local AI Inference |
| 7 | different-ai/openwork | +3,601 | 21,018 | Open-Source Claude Cowork |
| 8 | kira-96/jcode | +3,294 | 15,991 | RAM-Efficient AI Harness |
| 9 | ego-lite/ego-lite | +2,633 | 8,523 | Browser State Sharing |
💡 RunPod recommendation: Gpu cloud for ai inference and training from $0.29/hr. Try RunPod →
Affiliate disclosure: we may earn a commission if you sign up via this link, at no extra cost to you.
💡 Cloudflare recommendation: Cdn, dns, ddos protection, and edge compute for any site. Try Cloudflare →
Affiliate disclosure: we may earn a commission if you sign up via this link, at no extra cost to you.