Building With Llms
This consolidates techniques from 60 product leaders who've shipped real AI features into a single reference Claude can pull from when you're building. Use it when you're stuck on prompt design, debating RAG architecture, or need to justify why you're spending time on evals. The advice is opinionated and practical: few-shot examples beat descriptions, context engineering matters more than prompt tweaking, binary evals beat rating scales, and retry your failures because LLMs are stochastic. Honestly, it's most valuable for the specific patterns it can suggest when you describe your problem, like how to structure decomposition prompts or when to layer multiple models instead of fighting with one.
- installs
- 1.7K
- GitHub stars
- 1.4K
- last updated
- Aug 21, 2026
Install
$ npx -y skills add refoundai/lenny-skills --skill building-with-llms --agent claude-codeInstalls into .claude/skills of the current project.
Files
This consolidates techniques from 60 product leaders who've shipped real AI features into a single reference Claude can pull from when you're building. Use it when you're stuck on prompt design, debating RAG architecture, or need to justify why you're spending time on evals. The advice is opinionated and practical: few-shot examples beat descriptions, context engineering matters more than prompt tweaking, binary evals beat rating scales, and retry your failures because LLMs are stochastic. Honestly, it's most valuable for the specific patterns it can suggest when you describe your problem, like how to structure decomposition prompts or when to layer multiple models instead of fighting with one.
Browse the repository on GitHubSponsors
Details
- Repository
- github.com/refoundai/lenny-skills
- Categories
- DevOps & CI/CD, Documentation, Code Review & Quality, AI & Agent Building
