DEV Community

Fenju Fu
Fenju Fu

Posted on

Developers Don't Need Another Agent Framework — They Need Your `.agents` Directory

Today's GitHub Trending reveals a shift that's been building for months: the agent ecosystem is moving from framework hunger to usage hunger.

Look at what's trending:

  • mattpocock/skills — literally a developer's .agents directory, open-sourced. 1,225 stars today. Not a framework. Not a library. Just: "here's what I actually use."
  • homeassistant-ai/ha-mcp — lets Claude, ChatGPT, and other AI assistants configure and automate Home Assistant. 790 stars today. Agents are controlling real devices now.
  • anthropics/skills — Anthropic's official Agent Skills repository. 1,679 stars today. The ecosystem is consolidating around "skills" as the unit of agent capability.

The pattern is clear. Developers have enough frameworks. What they don't have is: a clear picture of how someone else actually sets up and uses an agent day-to-day.

The Gap Between "Here's a Skill" and "Here's How I Use It"

Anthropic can publish an official Skills repo. Matt Pocock can share his .agents directory. But between "here's a skill definition" and "here's how I configured my assistant to use it for daily work" — there's a gap that tutorial content needs to fill.

This is exactly what iflytek/astronclaw-tutorial addresses.

What AstronClaw and Loomy Tutorials Cover

iflytek/astronclaw-tutorial (https://github.com/iflytek/astronclaw-tutorial) provides hands-on tutorials for two assistants in the Astron ecosystem:

  • AstronClaw — cloud-based AI assistant
  • Loomy — desktop AI assistant

Desktop AI assistant connected and ready for automation tasks

The tutorials aren't API documentation. They walk through real usage scenarios: setting up your assistant, connecting skills, automating desktop tasks, and building daily workflows that actually save time.

Think of it as the ".agents directory" approach applied to a full assistant platform — not abstract capability descriptions, but step-by-step "here's how you use this" content.

Pairing with Official Skills

The tutorials work hand-in-hand with iflytek/iFly-Skills (https://github.com/iflytek/iFly-Skills) — iFLYTEK's official skill collection covering speech recognition, OCR, translation, proofreading, and multimodal capabilities. Instead of writing skills from scratch, you plug official skills into your assistant setup and follow the tutorial to get them running.

iFly-Skills official skill collection overview

Why This Matters Now

The mattpocock/skills repo proves that developers will star anything that shows real usage — even if it's "just" a Shell directory. The ha-mcp repo proves that agents are moving into real environments (homes, browsers, desktops). The anthropics/skills repo proves that the skill ecosystem is mature enough for official backing.

What's missing is the connective tissue: tutorials that show how to go from "I have an assistant" to "my assistant does useful work for me every day."

That's the gap astronclaw-tutorial fills.

Check it out: https://github.com/iflytek/astronclaw-tutorial

Top comments (0)