☁️ Cloud & Databases

Claude Code Skills: Great Recipes, Lousy Always-Ons

Skills in Claude Code promise smarts, but they flop because models ghost them. Time to shove best practices into CLAUDE.md where they actually stick.

Claude Code terminal showing SKILL.md files and invocation failures

⚑ Key Takeaways

  • Skills match CLAUDE.md quality when invoked, but activation fails 34-94% of the time. 𝕏
  • Use CLAUDE.md for best practices; reserve skills for action-specific recipes. 𝕏
  • Vercel's single-shot evals mislead β€” multi-turn reality favors persistent context. 𝕏
Published by

theAIcatchup

Community-driven. Code-first.

Worth sharing?

Get the best Open Source stories of the week in your inbox β€” no noise, no spam.

Originally reported by Dev.to

Stay in the loop

The week's most important stories from theAIcatchup, delivered once a week.