Post

I built a Claude Code plugin that teaches it how to work with Ruby - picks the right version manager, reads the right docs, connects to Ruby LSP. I think we will eventually have community maintained bridges between AI agents and the language ecosystem.

24/1/26, 2:36 pm4 reposts10 likes

Stan Lo

@st0012.dev

24 Jan

Here's the blog post: st0012.dev/2026/01/24/r...

st0012.devRuby Skills: Teaching Claude Code About Ruby's Tooling And EcosystemRuby developers using Claude Code hit this pattern: you ask it to run tests, it picks the wrong Ruby version, fails, tries bundle install, fails again.

Stan Lo

@st0012.dev

24 Jan

Repo: github.com/st0012/ruby-...

github.comGitHub - st0012/ruby-skills: Claude Code plugins for Ruby development.Claude Code plugins for Ruby development. Contribute to st0012/ruby-skills development by creating an account on GitHub.

Stephen

@diaconou.com

24 Jan

Nice! I’ll compare notes on the RBS typing part compared to my attempt at github.com/stevegeek/cl...

github.comGitHub - stevegeek/claude-ruby-plugins: Claude Code skills, commands, and agents for writing Ruby RBS type signaturesClaude Code skills, commands, and agents for writing Ruby RBS type signatures - stevegeek/claude-ruby-plugins

Stan Lo

@st0012.dev

24 Jan

Did you find CC doing better using the multi-agent (role) model? I haven't tried it so am just curious

Stephen

@diaconou.com

24 Jan

I’ve found subagents to be powerful as they have their own context window and so tend to work well performing a very specific “focussed” task. Also it’s cool that you can then ask CC to spawn multiple (it will spawn up to 10 in parallel)

Stephen

@diaconou.com

24 Jan

The main CC agent coordinates and collects the results but without the pollution of the exact implentstion details of each agents work, so good for large cross cutting tasks, eg adding types across a whole gem say