惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

A
About on SuperTechFans
博客园 - 聂微东
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
博客园 - 司徒正美
宝玉的分享
宝玉的分享
美团技术团队
量子位
The Cloudflare Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
IT之家
IT之家
爱范儿
爱范儿
J
Java Code Geeks
博客园 - Franky
Last Week in AI
Last Week in AI
B
Blog
H
Hackread – Cybersecurity News, Data Breaches, AI and More
I
InfoQ
GbyAI
GbyAI
Recent Announcements
Recent Announcements
小众软件
小众软件
H
Help Net Security
Microsoft Azure Blog
Microsoft Azure Blog
MyScale Blog
MyScale Blog

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
Chris Banes' Skills Repo: Claude Code Meets Android Devel...
Alan West · 2026-05-17 · via DEV Community

Last week Chris Banes — yeah, that Chris Banes from the Android team at Google — published a public repo of Claude Code skills targeting Kotlin, Jetpack Compose, and Android. I spent an evening poking around it, and there's enough here worth writing about.

I'm not going to pretend I've battle-tested every skill in the repo. But I've been using Claude Code skills for backend work for a few months now, and seeing someone with deep Android expertise share their setup publicly is a useful signal worth unpacking.

What Are Claude Code Skills, Anyway?

Quick refresher before we get into the repo itself. Skills are reusable instruction files you drop into a .claude/skills/ directory. Claude picks them up automatically and invokes the relevant ones when their description matches what you're trying to do. The official docs cover the format in more detail.

A typical skill file structure looks like this:

---
name: my-skill-name
description: "When to use this skill  Claude reads this to decide"
---

# Skill instructions
Step-by-step guidance for the model...

Enter fullscreen mode Exit fullscreen mode

That's it. Frontmatter with name and description, then plain markdown instructions. Easy to write, easy to share.

Why This Beats Pasting Snippets

Before I started using skills, the way I shared "how we do things here" with Claude was either a CLAUDE.md file or just pasting context at the start of every session. Both work, but they have tradeoffs.

CLAUDE.md gets loaded every time, which bloats context for tasks where the conventions don't apply. Pasting context manually means you forget half of it. Skills sit somewhere in the middle — they're available but only pulled in when relevant.

For Android work specifically, this matters more than you'd think. Compose has at least three reasonable ways to handle state depending on context, and getting Claude to pick the right one without explicit guidance is hit or miss.

A Practical Example

Say I'm scaffolding a screen with a list and detail view. Without guidance, Claude might give me back a perfectly valid MutableStateFlow in a ViewModel when what I actually wanted was a rememberSaveable because the state should survive config changes but doesn't need to live anywhere else.

A skill file for this might look like:

---
name: compose-state-decisions
description: Use when scaffolding Compose UI state — guides choice between rememberSaveable, ViewModel state, and remember
---

When picking state holders in Compose:
- `remember`: ephemeral UI state (expanded/collapsed)
- `rememberSaveable`: state that should survive config changes
- ViewModel state: state shared across screens or surviving process death

Prefer `rememberSaveable` over ViewModel for screen-local state
unless the state needs to be observed elsewhere.

Enter fullscreen mode Exit fullscreen mode

That's the kind of thing where a skill earns its keep. You're not writing code — you're encoding judgment that Claude can apply consistently across sessions.

Setting It Up

Cloning into your project's skills directory is the obvious first move:

# From your Android project root
mkdir -p .claude/skills
cd .claude/skills

# Pull in the repo (or copy specific skills you want)
git clone https://github.com/chrisbanes/skills.git chrisbanes

Enter fullscreen mode Exit fullscreen mode

After that, Claude Code should pick them up on the next session. You can verify by asking Claude what skills it has available.

One thing I'd recommend: don't just dump the whole thing in and forget about it. Read through the skill descriptions, decide which ones match how your team actually works, and prune the ones that don't fit. Skills you don't use are noise that can muddle relevance scoring.

Where This Pattern Has Legs

The interesting thing about Chris's repo isn't really the specific skills — it's that someone with serious Android credentials is sharing them publicly. That means others can fork, adapt, or just learn from how to structure skills well.

I'd love to see more of this across other ecosystems. A skills repo for Rust async patterns. One for Spring Boot conventions. One for React Server Components gotchas. The model is portable, and the more reference examples there are, the better everyone's skills get.

A few weeks ago I started writing skills for our team's internal conventions, and the format Chris uses is tighter than what I'd been doing. Short descriptions, clear when-to-use guidance, minimal prose. I'm going to refactor mine this week.

Things to Watch Out For

A few things I've bumped into running skills in real projects:

  • Skill descriptions matter more than the body. Claude uses the description to decide whether to invoke a skill. Vague descriptions equal skills that never fire.
  • Overlapping skills cause weird behavior. If two skills both claim to handle Compose state, you'll get inconsistent invocation. Be ruthless about scope.
  • Project-specific beats generic. A skill that says "use clean architecture" is too broad to be useful. One that says "use our Result<T> type for network calls" lands much better.

Also a small tooling aside — if you're tracking how often skills get used across your team (or just want analytics on your dev blog without the GDPR cookie banner dance), privacy-focused options like Umami or Plausible give you full data ownership and a much lighter footprint than Google Analytics. I migrated two side projects to Umami last year and haven't looked back.

Should You Use This Repo?

If you write Android, Kotlin, or Compose code and you're already using Claude Code, yes — at least take a look. Even if you don't adopt the skills wholesale, the patterns in how they're written are worth learning from.

If you're not using Claude Code yet for Android work, this repo is a reasonable nudge to try it. Pair it with your own CLAUDE.md and you've got a setup that respects your conventions without constant babysitting.

I haven't tested this thoroughly enough on a real feature to make grand claims — I'll probably write a follow-up once I've used it for a couple of weeks on something nontrivial. But the early signal is positive. The skill format is simple enough that even if Chris's repo isn't a perfect fit for your project, it gives you a clear template to write your own.

Go read the repo: github.com/chrisbanes/skills.