A Calude skill to build Claude skills?
I’d like to turn my own frameworks into Claude Code skills, not just prompt templates, but skills that preserve the original reasoning structure, questioning flow, prioritization logic, and behavioral consistency. What’s the best process for doing this properly? A few questions for experienced builders: 1. How do you structure complex frameworks into reliable Claude skills? 2. Is it better to use one large orchestrator skill or multiple modular skills? 3. How do you preserve deep reasoning instead of getting generic AI responses? 4. Any proven patterns for coaching, questioning, or diagnostic frameworks? 5. How do you handle sequencing, memory, and trigger conditions? 6. Any good repos, starter kits, or “meta-skills” for building custom skills? 7. What are the biggest mistakes people make when designing Claude skills? Example: One of my frameworks requires the AI to first diagnose the user’s active identity pattern before asking solution-oriented questions. The order matters a lot. Would love to learn from people already building advanced Claude Code skills at scale. Thanks heaps in advance for your takes!