smart-skill-loading

Smart skill loading auto-detects and loads only minimal metadata initially, then full content for matched tasks, reducing system prompt tokens by 1,000–1,400 per prompt. This progressive disclosure prevents context bloat while maintaining accuracy.

Progressive disclosure loads skill metadata first and fetches complete skill content only when relevant to your task. This approach trims unnecessary context, lowers costs, and speeds up AI responses without sacrificing capability matching.

Yes. Smart skill loading uses task analysis and skill matching to conditionally load only skills applicable to your request. For example, asking about PDF extraction loads only the PDF skill, not unrelated skills.

By loading only necessary skills, progressive disclosure eliminates irrelevant context that can confuse the model, improving task accuracy. Fewer tokens mean faster processing and lower API costs—up to 1,400 tokens saved per prompt.

Use smart skill loading when managing large skill sets, operating under strict token budgets, preventing context bloat, or ensuring only relevant expertise is injected for each task. It's automatic in multi-skill workflows.