Prompting

This is the meta layer for all prompt engineering in PAI. It owns templates, standards, and composition tools so other skills can generate prompts programmatically instead of duplicating patterns everywhere. The Handlebars-based system cut 35,000 tokens across PAI by separating structure from content. You'd use this when building dynamic agents, rendering workflow steps from YAML, or generating eval judges. The real value is treating prompts like code: templates are structure, data is content, and you compose them at runtime. It references 1,500+ papers and Anthropic's latest practices, but the practical win is that when you need to generate a briefing or validation gate, you call RenderTemplate instead of writing another 2,000 token blob.

Install to Claude Code

npx -y skills add danielmiessler/personal_ai_infrastructure --skill Prompting --agent claude-code