AI Repo: p-e-w/heretic

Skyler: Hey everyone, welcome back to Git News AI, your go-to source for the hottest A I repositories blowing up on GitHub right now. Big thanks to our sponsors today — Ada, who helps teams integrate A I seamlessly into their workflows, Ago Consulting, your silicon development partners from A I all the way to SoC, and Zen Semiconductor, an umbrella company with a whole portfolio of silicon-industry companies, plus Ada for deploying A I at scale in any business, including their Vect appliance which runs local L L Ms and vector search on dedicated hardware. Skyler: Today we've got five repos that are genuinely turning heads — we're talking censorship removal for language models, a seriously powerful OSINT suite, an open source S E O platform, a multi-agent classroom experience, and a massive curated list of Claude A I skills. Skyler: Let's get into it. Skyler: Alright, kicking things off with one that I guarantee is going to spark some conversation — it's called Heretic, from the GitHub user p-e-w, and it has nearly twenty-nine thousand stars. That's not a small number, people. Skyler: So what does Heretic do? The description is pretty blunt — it's fully automatic censorship removal for language models. Yeah. That's the whole pitch. And honestly, the community has responded loudly to that. Skyler: The idea here is that modern L L Ms come with a lot of guardrails baked in, and while those guardrails serve real purposes, there are plenty of legitimate use cases — researchers, security professionals, fiction writers, red teamers — where those restrictions get in the way. Skyler: Heretic is a Python-based tool that systematically tests and works around those restrictions in an automated way. It's not a jailbreak prompt you paste in manually — it's more sophisticated than that. It runs through tests and now, as of the recent activity, it even shows you a summary at the end of your test runs, which is a nice quality-of-life touch. Skyler: They just bumped to version two-point-zero-dev, which tells me this project is actively evolving and the maintainers are serious about it. This isn't a weekend experiment someone abandoned. Skyler: Now I want to be clear — this kind of tool lives in ethically nuanced territory. Using it responsibly matters. But from a pure research and red-teaming standpoint, the technical work here is impressive. Understanding where your model's restrictions are is genuinely useful if you're building on top of L L Ms or doing safety research. Skyler: If you're an A I safety researcher, a developer building products on top of language models who needs to understand the boundaries, or just someone philosophically interested in how L L M censorship actually works under the hood — Heretic is worth a look. Skyler: Twenty-nine thousand stars don't lie. Check the show notes for the link. Skyler: Next up we've got user-scanner from kaifcodec, and this one is sitting at about three-and-a-half thousand stars, but don't let that fool you — the scope of this tool is wild. Skyler: The tagline calls it a two-in-one Email and Username OSINT suite for deep data extraction. OSINT, for anyone not familiar, stands for Open Source Intelligence — basically, gathering information from publicly available sources. Skyler: And what makes user-scanner stand out is the sheer number of scan vectors it covers. We're talking over four hundred and fifty actively maintained vectors — more than a hundred and seventy-five for email addresses and over two hundred and eighty for usernames. That is a lot of surface area. Skyler: So the core idea is you drop in a single email or username and this tool fans out across hundreds of platforms and data sources to build a digital profile. It's designed for security research, investigations, and digital footprinting — which are all legitimate and important use cases in the security world. Skyler: The recent activity includes adding something called the ASL Bloom module in the learning category under email scanning. I love that they're adding educational context to the scans — not just "here's data" but helping users understand what they're looking at. Skyler: The A I relevance here is real — tools like this increasingly feed into A I pipelines for threat intelligence, fraud detection, and identity verification systems. If you're building anything in the security A I space, this is the kind of data collection layer that plugs right in. Skyler: If you're a security researcher, a digital forensics person, or you're building A I-powered threat intel tools, user-scanner is absolutely worth exploring. And as always, use it responsibly and legally. Skyler: Check the show notes for the link. Skyler: Okay, repo number three is open-seo from every-app, and this one has nearly fifteen thousand stars and a really clear value proposition — it's an open source alternative to Semrush and Ahrefs. Skyler: If you've ever tried to get a Semrush or Ahrefs subscription, you know those tools are expensive. Like, really expensive. So an open source alternative that covers that same territory is genuinely exciting news for indie developers, small teams, and startups. Skyler: This is TypeScript-based, so it's web-native and feels like something you could self-host or integrate into an existing stack without too much friction. The fact that it's already at fifteen thousand stars suggests the developer community agrees there's a real gap here. Skyler: Now, where does A I come in? The recent activity is interesting — they added a Cursor marketplace plugin package and updated ChatGPT app submission metadata. So they're actively building out integrations with A I coding tools and ChatGPT, which positions this not just as a traditional S E O tool but as something that's evolving to work inside A I workflows. Skyler: Think about it — if you're using an A I agent to manage content strategy or run automated site audits, having an open source S E O engine you can actually hook into is huge. You're not locked into a third-party API with rate limits and pricing tiers. Skyler: For developers building content tools, A I agents that work with web data, or just anyone doing S E O work who doesn't want to pay enterprise prices, open-seo deserves a spot on your radar. Skyler: Links in the show notes as always. Skyler: Alright, this next one really caught my eye — it's OpenMAIC from THU-MAIC, which stands for Open Multi-Agent Interactive Classroom, and it's got over twenty-two thousand stars. This is a serious project. Skyler: The pitch is simple and kind of brilliant — get an immersive, multi-agent learning experience in just one click. And when they say multi-agent, they mean it in the true A I sense. Multiple A I agents interacting with each other and with you to create a dynamic classroom environment. Skyler: This is coming out of Tsinghua University, which if you know anything about A I research, is one of the top institutions in the world for this stuff. So the academic credibility is there. Skyler: What makes this different from just, say, an A I tutor chatbot? The multi-agent aspect means you can have different agents playing different roles — a teacher agent, maybe debate opponent agents, peer student agents — all interacting in real time to create something that feels more like an actual learning environment than a one-on-one Q&A session. Skyler: The recent performance improvements are meaningful too — they sped up classroom loading through better media hydration and sidebar thumbnails, and they fixed how intermediate message frames are handled in storage. These are the kinds of optimizations that show a team thinking seriously about real-world usability at scale. Skyler: TypeScript-based, so the frontend is clean and deployable, and clearly there's a lot of thought going into the U X here. Skyler: If you're an ed-tech developer, a researcher exploring multi-agent systems, or honestly if you just want to see what the cutting edge of A I-powered learning looks like, OpenMAIC is a must-check. The fact that it's open source means you can take the architecture and learn from it even if you're building something totally different. Skyler: Show notes have the link. Skyler: And okay, saving one of the biggest for last — awesome-claude-skills from ComposioHQ. This thing has seventy-three thousand, nine hundred and twenty-six stars. Let that sink in. Skyler: So this is a curated awesome-list style repo — if you've been around GitHub for a while you know the awesome-list format — it's a community-maintained collection of resources, tools, and examples around a specific topic. And this one is all about Claude skills, resources, and tools for customizing Claude A I workflows. Skyler: Claude, for anyone who needs the quick intro, is Anthropic's flagship L L M, and it's been gaining serious traction as a developer-first model. It's known for being thoughtful, long-context capable, and genuinely useful for complex reasoning tasks. Skyler: What ComposioHQ has done here is aggregate the best skills, integrations, and customization patterns for Claude into one place. We're talking about ways to extend Claude's capabilities, plug it into different tools and workflows, and get more out of it for specific use cases. Skyler: The recent activity is just README updates, which for a repo like this is actually exactly what you want to see — it means the community is continuously adding new skills and resources. An awesome-list lives and dies by how fresh its content is. Skyler: ComposioHQ itself is known for building agent infrastructure, so they have real credibility curating something like this. They know what actually works in production agent pipelines. Skyler: If you're building with Claude — and given its growing adoption, there's a real chance you are or soon will be — this is basically your cheat code. Instead of spending weeks discovering tools and patterns through trial and error, this repo gives you a running start. Skyler: Seventy-three thousand stars. That's a community saying this resource is genuinely valuable. Don't sleep on it — link is in the show notes. Skyler: Alright, that's a wrap on today's episode of Git News AI — five repos, a ton of ground covered, and hopefully a few new things to dig into this week. Skyler: Massive thank you to our sponsors — Ada for helping teams actually get A I into their workflows, Ago Consulting for expert guidance from A I all the way down to silicon, and Zen Semiconductor for pushing what's possible in hardware A I deployment, including their Vect appliance for running local L L Ms and vector search right on the device. Skyler: If you want more great content from the Ada podcast network, definitely check out AI Special Report — it's sharp, it's timely, and it goes deep on the A I stories that matter. Subscribe to Git News AI wherever you get your podcasts, and as always, all the links from today's episode are waiting for you in the show notes. See you next time.