<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Skills on Raphaël Bellec</title><link>https://rbellec.github.io/blog/en/tags/skills/</link><description>Recent content in Skills on Raphaël Bellec</description><generator>Hugo -- gohugo.io</generator><language>en</language><copyright>© 2026 Raphaël Bellec</copyright><lastBuildDate>Wed, 29 Apr 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://rbellec.github.io/blog/en/tags/skills/index.xml" rel="self" type="application/rss+xml"/><item><title>A Claude Code skill that ships, tested across four games</title><link>https://rbellec.github.io/blog/en/posts/skill-claude-code-bga/</link><pubDate>Wed, 29 Apr 2026 00:00:00 +0000</pubDate><guid>https://rbellec.github.io/blog/en/posts/skill-claude-code-bga/</guid><description>&lt;p>How do you build a skill that actually ships? How do you build a test loop that closes? What does an agent decide on your behalf when the spec stays silent? Adapting board games on Board Game Arena gave me a case study — bounded scope, constrained framework, instant validation. One example: April 10, 2026, with the help of this skill, I let Claude start from a game&amp;rsquo;s rules (Quantum Tic-Tac-Toe) and reach a complete game in under three hours, without writing a single line of code by hand. Four games later, here&amp;rsquo;s what I took away.&lt;/p></description></item></channel></rss>