AI Repo: anthropics/skills

Skyler: Hey everyone, welcome back to Git News A I — the show where we dig into the hottest A I repositories trending on GitHub so you don't have to scroll through the noise yourself. Big thanks to our sponsors today: Ada, helping teams integrate A I into their workflows seamlessly; Ago Consulting — that's "Ahh-Go" — your go-to for silicon development from A I all the way down to SoC design; 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 — and their Vect appliance is seriously cool, running local L L Ms and vector search on dedicated hardware. Alright, today we've got five repos on the docket — we're talking Agent Skills from Anthropic, a Go guidelines repo from JetBrains built for A I coding agents, the legendary Awesome ChatGPT Prompts reborn, a Chinese stock screening system, and a tool that turns your entire codebase into a queryable knowledge graph. Let's get into it. Skyler: Okay, first up — and this one has my full attention — it's anthropics/skills, the public repository for Agent Skills from Anthropic themselves. This thing is sitting at a staggering 173,000-plus stars on GitHub, which tells you everything about how much attention the developer community is paying to it. So what is it exactly? Agent Skills is essentially a curated, open collection of skills — think modular capabilities — that you can plug into A I agents built on top of Claude. We're talking reusable building blocks that let agents do real things: design frontends, interact with APIs, manage costs, orchestrate complex workflows, and more. The recent activity tells a great story here — they just pushed an update to the frontend design skill specifically to avoid those frustrating generic design defaults that A I tends to fall into. You know what I mean — every A I-generated UI looking like the same Bootstrap template. This is a real pain point they're actively solving. They also updated the Claude A P I skill to support Claude Fable 5.1 and Mythos 5.1 — new model variants that are apparently baked right into the skill definitions. And there are Managed Agents updates too, with a focus on cost optimization, which honestly is one of the most underrated concerns when you're running agents at scale. If you're building anything on top of Claude — whether it's a coding assistant, a research agent, a workflow automation tool — this repo is essentially your skill library. Think of it like npm, but for agent capabilities. M L engineers and agent developers, this one is a must-star. Skyler: Next up we've got something a little more niche but genuinely interesting — JetBrains/go-modern-guidelines. Now, JetBrains you probably know as the folks behind IntelliJ, GoLand, and a whole suite of developer tools. This repo sits at around 3,100 stars and the premise is delightfully simple: help A I coding agents write modern Go. That's it. That's the pitch. And honestly? It's kind of brilliant. Here's the problem it's solving — A I coding assistants, whether that's GitHub Copilot, Cursor, or whatever L L M-backed tool you're using, they tend to generate Go code that's syntactically correct but stylistically outdated. Go has evolved a lot, and the training data for most models skews toward older patterns. So JetBrains is publishing a living document of modern Go idioms, guidelines, and best practices specifically formatted and structured so that A I agents can consume and apply them. It's less of a codebase and more of a knowledge base — but it's versioned, community-contributed, and actively maintained. They just added a CONTRIBUTING.md which signals this is meant to be a community effort, not just a JetBrains internal doc that got open-sourced. If you're a Go developer who's frustrated by your A I assistant writing Go like it's 2017, or if you're building A I tooling that generates Go code, this repo is worth bookmarking. It's a small but thoughtful contribution to the ecosystem. Skyler: Alright, repo number three — f/prompts.chat, formerly known as Awesome ChatGPT Prompts. If you've been in the A I space for more than five minutes, you've probably encountered this one. It's sitting at nearly 169,000 stars and it's one of those foundational community resources that just keeps growing. So what's new? Well, the project has evolved beyond just a curated list — it's now a full platform where you can share, discover, and collect prompts from the community. It's free and open source, and crucially, you can self-host the whole thing for your organization with complete privacy. That last part is huge for enterprises who want the value of a shared prompt library without their prompts leaking to a third party. The recent commits are a great snapshot of how global and diverse this community is — there's a new prompt for analyzing Chinese stock trends, and another for assessment workflows. That breadth is kind of the whole point. This isn't a curated top-ten list — it's a living, breathing library of prompts for every use case imaginable. If you're a developer building prompt-heavy applications, a data scientist trying to get better outputs from your L L Ms, or even a non-technical user who just wants to get more out of ChatGPT, this is essential. And if you're running a team that's standardizing on certain prompts internally, the self-hosting option makes this genuinely enterprise-ready. Check the show notes for the link. Skyler: Okay, shifting gears a bit — repo four is sngyai/Sequoia-X. This one is a bit more specialized but super interesting from a real-world A I application standpoint. Sequoia-X is an automated stock selection system for the Chinese A-share market. We're talking about a fully automated pipeline that scans for multiple technical chart patterns, runs after market close, and then pushes notifications directly to Feishu — which is ByteDance's enterprise communication platform, kind of like Slack but huge in China. It's written in Python and has about 6,500 stars, which is solid. Now, the A I relevance here is more on the applied automation side — this isn't a deep learning model predicting stock prices, but it is a great example of using programmatic pattern recognition and automated agents to do work that humans would otherwise have to do manually, every single day after the bell rings. The technical pattern scanning is the core of it — think classic chart analysis: support and resistance levels, breakout patterns, volume signals, all automated. And the Feishu integration means you get your results pushed to you without having to log in anywhere. There haven't been recent commits, so this might be more of a stable, mature tool than an actively evolving one, but the codebase is there and it's open source. If you're a quant developer or a finance-adjacent engineer in the Chinese market ecosystem, or honestly if you just want to study how someone built an end-to-end automated financial analysis pipeline in Python, this is worth a look. Skyler: And rounding out today's episode — saving arguably one of the most technically exciting for last — we have Graphify-Labs/graphify. This one is wild. 114,000-plus stars and it's doing something really clever. The pitch: turn any codebase, along with its docs, S Q L schemas, config files, and P D Fs, into a queryable knowledge graph. And it comes as a slash-graphify skill for Claude Code, Cursor, Codex, and Gemini C L I. So it plugs directly into the A I coding tools you're probably already using. Here's why this matters. Most A I coding assistants rely on vector search and embeddings to find relevant context in your codebase — which works okay, but it's fuzzy. Graphify takes a completely different approach: local, deterministic A S T parsing. A S T stands for Abstract Syntax Tree — it's a structured, precise representation of your code. So instead of semantic similarity, you get actual relationships — this function calls that function, this class inherits from this base, this config value is consumed here. Every edge in the graph is explained, meaning you can actually understand why two things are connected. And there's no vector store required, which means no cloud dependency, no embedding costs, no hallucinated connections. It runs entirely locally. The practical upside? When you ask your A I coding agent a question about your codebase, it's working from a precise, structured map of your code rather than a probabilistic guess. For large codebases with complex dependencies, this could be genuinely transformative. Engineers working on big legacy systems, senior developers trying to onboard A I agents into complex repos — this is the tool you want to look at. Definitely check the show notes for the link on this one. Skyler: Alright, that's a wrap on today's repos — what a lineup. Huge thank you to our sponsors: Ada for helping teams put A I to work in real workflows, Ago Consulting — "Ahh-Go" — for expert silicon development guidance from A I chips all the way to full S o C design, and Zen Semiconductor for their incredible portfolio of silicon companies and the Vect appliance bringing local L L M and vector search power to any organization. If you're hungry for more A I and tech coverage, go check out AI Special Report — another great show in the Ada podcast network — it's packed with sharp takes on where the industry is headed. And hey, if you're not subscribed to Git News A I yet, now's the time — new episodes drop regularly and all the repo links we talked about today are waiting for you in the show notes. See you next time.