GitHub Trends · Nov 10: Two Agent Skill Libraries Go Head-to-Head

There are a bunch of new faces on GitHub Trending today. And look — right out of the gate, several heavy hitters.

I scanned the list. Honestly, today's wave is pretty interesting — two main threads. One is skill libraries for AI coding agents, and the other is an old-school infrastructure project suddenly making a comeback.

Let's start with the hottest one. addyosmani's agent-skills gained over 1,100 stars in a day, and it's already at over 76,000 total stars.

The name says it all: Production-grade engineering skills for AI coding agents. In other words, a production-grade engineering skill pack for coding agents.

Wait, isn't this kind of like that anthropics/skills project? That one has over 160,000 stars.

Great point. They're complementary — maybe even a bit of a rivalry. Addy Osmani is a veteran from the Google Chrome team, and his approach leans more toward engineering practice, toward real-world application.

So what skills are actually in there? Is it like prompt templates for writing code?

Not just templates. It's more like taking engineering standards from real projects — how to review code, how to write tests, how to handle refactoring — and breaking them all down into forms an agent can directly call.

Notice it's written in JavaScript, which means many of the skills exist as executable scripts. It's not just talk — it's the kind of thing an agent can actually run.

Oh, I get it. It's like an employee handbook for new hires, but every item in the handbook comes with an automation script.

Exactly, that's the feel. And I think there's a backdrop to why it blew up today: people are realizing agents are great at writing small demos but fall flat the moment they hit a real project. What's missing is exactly this kind of engineering discipline.

So it fills a gap. Who's it for? I'd say not just programmers.

Right, product managers can use it too. Think about it — when you ask Claude Code or Cursor to change something and you've got this skill set installed, the code quality and consistency of what it outputs will be much more stable.

To sum up: agent skill libraries are moving from official examples to community实战 — real-world combat. Engineering落地 — making it real — is today's deciding factor.

Okay, thread two: anthropics' own claude-quickstarts also made the list. It's not gaining fast — only 28 stars in a day — but it's nearing 20,000 total.

Yeah, what's interesting is it's not a new project — it's an official set of quick-start templates. Making the list today is more like being swept up by the agent-skills hype.

So how is it different from agent-skills? Both teach you how to use Claude.

Quickstarts teaches you how to hook the Claude API into a deployable app. It's more introductory, more zero-to-one. Agent-skills teaches agents already writing code how to do it more professionally. More one-to-hundred.

That connects perfectly. Newcomers check out quickstarts, veterans top up with agent-skills — a one-two combo.

Right, and it's in Python, the friendliest ecosystem, low barrier to entry. Both appearing today — I don't think it's a coincidence.

What does it mean?

It means Claude's developer ecosystem is shifting from 'can it be used' to 'is it being used properly.' That's a sign of maturity.

Then there's thread three, and the style completely changes. we-promise/sure — a personal finance app written in Ruby — gained over 180 stars in a day.

Yeah, it really stands out among all the AI projects. Its positioning is 'for everyone, by everyone' — meaning personal finance shouldn't be only for people who understand finance.

Now I'm curious. There are tons of finance apps out there — why did it suddenly blow up today?

I think two reasons. One, it's open source and self-hosted — your bill data doesn't go to any cloud provider. Two, it's written in Ruby, which is super friendly for people who like to tweak code for their own use.

Hey, I'm familiar with this pattern. Remember daily_stock_analysis, where forks outnumbered stars — lots of users just modified the code for themselves. Looks like it's a trend.

Right, open source plus self-hosting is becoming a core selling point for personal data apps. People don't trust the cloud anymore — they trust the code in their own hands.

Next up, tokio-rs/topcoat. Only 1,200-odd stars, but it gained over 430 in a day — the growth rate is scary.

It's a Rust web framework, and its motto is batteries-included — everything's installed for you, works out of the box. That's a pretty big shift in the Rust world.

Why do you call it a shift?

Because Rust traditionally values minimalism — you assemble everything yourself. Topcoat goes the opposite way, like Go or Rails, configuring all the common features for you.

Oh, like the difference between a finished apartment and a bare shell.

The finished-apartment analogy is spot on. And it's from the official tokio team — official backing is exactly why it's climbing the charts today.

Last one, google-deepmind/mujoco. Over 16,000 stars, up 26 in a day — doesn't look like much, but it's an old-school physics simulator.

MuJoCo is a real veteran — Multi-Joint dynamics with Contact, infrastructure for robot control and reinforcement learning. It also got a boost last month thanks to NVIDIA.

So what does its making the list today tell us? That old projects can get dug up too?

It tells us the embodied AI hype is spilling over. When people do robotics or reinforcement learning, they can't avoid physics simulation, can't avoid MuJoCo. It's on the list because demand pushed it there.

Let's sum up today's five. Two agent skill libraries, one open-source finance app, one Rust framework, one physics simulator.

If you ask me, the common thread in one sentence: they're all solving real-world problems. No fluff — all digging into engineering practice and real scenarios.

My must-sees today: two. addyosmani/agent-skills — anyone doing agents has to take a look. Second is topcoat — if you want to save effort on Rust backends, check it out.

I'll add sure. If you care about personal data sovereignty and want to control your own bills, it's worth a try.

That's it for today. We'll keep watching the charts tomorrow. If new faces show up, we'll talk about them right away.