AI Repo: HKUDS/AI-Trader

Skyler: Hey everyone, welcome back to Git News A I, the show where we dig into the hottest A I repositories blowing up on GitHub so you don't have to spend hours doom-scrolling through trending pages yourself. Skyler: Today's episode is brought to you by LimitLess A I — more on them in a bit, but seriously, they're doing some cool stuff worth checking out. Skyler: Alright, we've got a packed lineup today — fully automated trading agents, a massively popular agent framework, a treasure trove of Claude skills, official Claude plugins, and a privacy-first healthcare A I tool that I'm genuinely excited to talk about. Let's get into it. Skyler: First up, we've got AI-Trader from HKUDS, sitting at over twenty thousand stars on GitHub, and the tagline alone is enough to raise some eyebrows — "100% Fully-Automated Agent-Native Trading." So what does that actually mean? Skyler: Essentially, this is a Python-based framework that lets A I agents take the wheel on financial trading — end to end, no human intervention required during execution. We're talking about an agent-native architecture, which means the whole system is designed from the ground up around autonomous agents rather than bolting A I onto some legacy trading logic. Skyler: What makes this interesting from a technical standpoint is that it's not just a wrapper around some trading API. The project is built to support the full trading pipeline — market analysis, decision making, order execution, all of it handled by the agent. Skyler: Now, worth noting — the repo hasn't seen recent commits, so it might be in a more stable or maintenance phase rather than active development. But with over twenty thousand stars, clearly the community has found real value here. Skyler: If you're an M L engineer who's been curious about applying reinforcement learning or L L M-based reasoning to financial markets, this is a fantastic starting point to understand what an agent-native approach actually looks like in practice. Skyler: Obviously, please, please don't just plug this into a live trading account without understanding what's happening under the hood. But as a learning resource or a foundation to build on? Really compelling stuff. Skyler: Who should check this out? Developers interested in agentic systems, quant-curious engineers, or anyone who wants to see how autonomous A I agents can be applied outside of the typical chatbot use case. Skyler: Alright, check the show notes for the link, and let's move on to something absolutely massive. Skyler: Repository number two is hermes-agent from NousResearch, and folks, this one is sitting at nearly two hundred and eleven thousand stars. Let me say that again — two hundred and eleven thousand stars. That is enormous. Skyler: The description is simple but evocative — "The agent that grows with you." And based on the recent activity, this project is very much alive and kicking. The team just shipped a WhatsApp dashboard pairing flow and fixed some session management behavior around dashboard chat resume, which tells me these folks are building toward a real-world, production-grade agent experience. Skyler: So what is hermes-agent at its core? It's a Python-based agent framework designed to be deeply adaptive — the idea is that this isn't a static tool, it evolves alongside how you use it and what you need from it. Skyler: The WhatsApp integration is a big signal here. We're moving past the era of A I agents that only live in a terminal or a web UI — hermes-agent is pushing into the communication channels people actually use every day. Skyler: The session management fixes are also worth paying attention to. Managing conversational state across sessions is one of the genuinely hard problems in building persistent agents, and it's great to see active work there. Skyler: NousResearch has a solid reputation in the open-source A I community — they've consistently pushed out high-quality models and tooling, so hermes-agent carries some serious credibility behind it. Skyler: If you're building production agents, experimenting with multi-channel A I assistants, or just want to study how a top-tier team architects an agent framework, this repository deserves serious attention. Skyler: With that star count, the community has clearly already voted. Check the show notes for the link — this one is a must-explore. Skyler: Okay, number three — awesome-claude-skills from ComposioHQ, clocking in at over sixty-seven thousand stars. This is one of those "awesome list" style repos, and it is packed. Skyler: If you're not familiar with the format, an awesome list is basically a curated, community-driven collection of resources, tools, and examples around a specific topic. And this one is laser-focused on Claude A I workflows and skills. Skyler: So what's a Claude skill? Think of it as a reusable capability or workflow that you can plug into Claude to extend what it can do — kind of like plugins but more customizable and often more developer-friendly. Skyler: The recent additions are telling — there's a new "overkill skill" which honestly I love the name, and a "building-blog skill" added under Development and Code Tools. That category alone tells you a lot about who's using this. Skyler: This repo is essentially a living directory of ways people are bending Claude to their will for real-world tasks. Code generation, content workflows, automation pipelines — it's all in there. Skyler: ComposioHQ has been doing great work in the A I tooling space, and curating this list shows real commitment to the Claude developer ecosystem. Skyler: If you're building with Claude — whether through the A P I or Claude Code — this is the kind of resource that can save you hours of reinventing the wheel. Someone has probably already built a skill for whatever you're trying to do. Skyler: Developers who want to supercharge their Claude workflows, prompt engineers, and honestly anyone building automation with Anthropic's models should have this bookmarked immediately. Show notes have the link. Skyler: Alright, repository four is closely related — claude-plugins-official from Anthropic themselves, with over thirty-one thousand stars. And yes, this is the official one. Skyler: Where awesome-claude-skills is community-driven, this repo is Anthropic-managed, which means everything in here has gone through some level of quality control and vetting. It's described as an "official directory of high quality Claude Code Plugins." Skyler: The recent updates include a Langfuse observability listing update and a database-on-AWS bump, which tells you something really interesting about what kinds of plugins are in here — we're talking about observability tooling, cloud infrastructure integrations, real developer infrastructure stuff. Skyler: Langfuse, if you're not familiar, is an open-source L L M observability platform. Seeing it in the official Claude plugins directory is a great sign for developers who care about tracing, monitoring, and debugging their A I applications. Skyler: The fact that Anthropic is actively maintaining this directory rather than just letting the community run wild is a maturity signal for the Claude ecosystem. They're investing in making Claude Code a platform, not just a product. Skyler: If you're using Claude Code in any professional or production context, you really want to know what's in this directory. These plugins can extend your Claude Code setup with capabilities that are tested, documented, and trustworthy. Skyler: Think of it like an app store but for Claude Code — except curated by the people who built Claude. That's a meaningful quality bar. Skyler: Engineers building internal tooling with Claude Code, teams evaluating A I coding assistants for enterprise use, and anyone who wants to get more out of Claude Code without building everything from scratch — this one's for you. Link is in the show notes. Skyler: And our final repository today is one that genuinely surprised me with how thoughtful it is — openmed from maziyarpanahi, with over four thousand stars and a really compelling mission. Skyler: The tagline is "Local-first healthcare A I" and it covers two big things — clinical named entity recognition, which is basically teaching A I to identify medical concepts in text, and HIPAA P I I de-identification, meaning it strips out patient-identifying information to keep data private. Skyler: And here's the part that really stands out — it runs one hundred percent on-device. No cloud, no patient data leaving your network. For healthcare, that's not just a nice-to-have, it's often a legal and ethical requirement. Skyler: The project supports over a thousand medical models, works across twelve languages, and runs on both Apple M L X and standard Python environments. That's an impressive range of support. Skyler: The recent activity shows active pull request merges around relation assertion and signature parsing, which suggests the team is working on deeper clinical reasoning capabilities — not just recognizing entities but understanding relationships between them. Skyler: In healthcare A I, the ability to say "this drug is prescribed for this condition" rather than just identifying the drug and the condition separately — that's a meaningful leap in clinical utility. Skyler: The Apache 2.0 license is also a big deal here. Open-source, commercially usable, and privacy-preserving? That's a combination that could make this genuinely adoptable in real healthcare settings. Skyler: If you're an M L engineer working in health tech, a data scientist dealing with clinical notes, or a developer building any kind of medical documentation tooling, openmed is absolutely worth your time. The on-device angle alone makes it stand apart from most solutions in this space. Skyler: Four thousand stars might sound modest compared to some of the other repos we covered today, but this is a specialized domain and the quality is clearly there. Check the show notes for the link. Skyler: And that's a wrap on today's episode of Git News A I — massive thanks to LimitLess A I for sponsoring the show, we genuinely appreciate the support and encourage you to check out what they're building. Skyler: If you found any of these repositories interesting, all the links are waiting for you in the show notes, so go explore, star some repos, maybe even contribute if something catches your eye. Skyler: Hit subscribe wherever you're listening, tell a developer friend about the show, and we'll see you next time with more of the freshest A I repos dropping on GitHub. Later, everyone!